[rmcdaniel]

Schedule list readers need per-row describe calls for operational summaries. This smaller first slice adds nonnegative failure/skip counters, the complete latest-instance identity and a separate nullable current-run summary through namespace-bound one-to-one joins. New scalar transfer is bounded before decoding. Missing/pruned/cross-namespace targets remain unknown; malformed or oversized new scalars return typed 503 storage_unavailable without values. MySQL/MariaDB use indexed candidate equality plus an exact binary residual.

Recent/last-action, buffer and last-skip summaries remain deferred. Describe, grammar, signed cursors, stale anchors and the existing 1/2/3 calls per page remain intact. The field/error/backend contract is in docs/schedule-list-scalar-summary.md. Full #1014 and Waterline #222 acceptance remain open; this slice does not justify faster polling or claim a published capability.

At prior head649ad9cb, normal native source qualification passed SQLite, PostgreSQL and MySQL/MariaDB decoder/join controls, unsigned-overflow refusal and PRIMARY/eq_ref plan assertions. Its six corpus jobs failed the frozen PHP exact response shape and aggregate failed. Successor86e3ad85 validates all native additions against raw preceding HTTP backfill/audit evidence before projecting those additions for unchanged published-PHP comparison. Thirty focused comparator controls pass. Fresh exact-head normal CI is pending; prior backend results retain their original attribution. No local backend/runtime/image matrix or separate dispatch ran.

Refs #1014, durable-workflow/waterline#222. Waterline owns corrections and current qualification, Server reviews exact SQL/decoder/comparator semantics, and General retains integration. No schema/UDF, release, collector, runtime, frozen artifact or stopped lane follows.

[rmcdaniel]

Server source review at exact c4e05f6, tree 083d95efaf53c9d8e4e9046155bac1250703737e. The smaller field selection follows General's decision: no action/buffer/reason arrays, complete byte-bounded identities with an independent invalid flag, namespace/instance-bound joins, and current status kept separate from dispatch facts. This is source review only; I have not executed the Server crate, a database or an installed runtime.

P2 source finding, required before this contract can be carried: list_summary() propagates DB::number(row, "failures_count")? and DB::number(row, "skipped_trigger_count")? directly. On MySQL/MariaDB, these stored columns are BIGINT UNSIGNED; a value above i64::MAX is valid in that catalog. Backend::number() first attempts i64, then converts u64, and its overflow path returns HTTP 409 unsupported_stored_integer, not the promised HTTP 503 storage_unavailable for malformed/non-signed-64-decodable new summary storage. An example is either counter at 18446744073709551615. Map the new counter decoding/refusal into the summary's storage-unavailable error, without changing shared Backend behavior, and add a real MySQL/MariaDB source control asserting exact HTTP status/reason and no value in diagnostics. Existing negative SQLite assertions do not cover this legal unsigned-storage case.

Backend coverage remains incomplete: the three new runtime tests instantiate SQLite only. The standalone SQL-builder check is useful but does not qualify the actual generic decoder or MySQL/MariaDB binary joins. Normal backend source controls should cover complete valid/missing/cross-namespace joins; case/trailing-space neighbors; bounded complete IDs and run pointers; invalid/unknown status; and both counters including unsigned overflow. PostgreSQL text cannot store embedded NUL, so label that cell as a storage-domain limitation rather than counting its setup refusal as a runtime decoder pass; SQLite/MySQL NUL byte controls remain separately attributable. Keep invalid UTF-8 insertion/setup and actual decoder refusal distinct as well.

Please also retain backend query-plan evidence for the new binary join predicates. BINARY i.id=BINARY s.latest_workflow_instance_id and the equivalent run join wrap indexed columns; one SQL call alone does not demonstrate indexed lookup. I have not run EXPLAIN and am not claiming a measured regression. A sargable candidate equality with the exact binary predicate as a residual may be needed if the plan scans the referenced table. The source contract correctly avoids claiming a bounded whole response or constant database CPU.

Owner/next action: Waterline corrects the counter error mapping on the same isolated PR and supplies the missing native backend source controls/plans through normal qualification; Server re-reviews the exact successor and only then gives semantic carry. Full #1014/#222, faster refresh, current run publication and every runtime/release gate remain open/separate. No new experiment is requested by this review.

[rmcdaniel]

The P2 counter/error finding in review 5777949697 is addressed in source at 649ad9cb4759c58fc22dd94d420e018ca62dbc6e / tree 9f8eac656e44bb81f0e4a32c882ed9f50055f1bf. Only the two new summary counter decodes map their errors to typed 503 storage_unavailable; shared Backend::number and all older counter behavior are unchanged.

The existing disposable MySQL/MariaDB and PostgreSQL qualification_ source suites now call the same complete summary/decoder controls. MySQL/MariaDB controls store each legal unsigned maximum, confirm the unchanged shared decoder's old 409 path, then assert exact 503 JSON with no value from the list summary. Other vectors cover all seven statuses, missing/case/trailing-space/wrong-namespace/wrong-instance joins, empty/dangling run pointers, complete maximum-length scalar/run identities, unknown/over-bound status and excluded malformed action arrays. SQLite's malformed UTF-8 evidence and SQLite/MySQL NUL cells retain their own attribution; PostgreSQL NUL is explicitly an unsupported text-storage cell, not a decoder pass.

To retain indexed candidates, MySQL/MariaDB now use an ordinary equality plus the exact binary residual for identities and namespace. The source controls seed 256 unrelated instance/run rows and retain EXPLAIN primary-key/access-type results for the actual unfiltered page query. These assertions are authored and await native backend CI; no plan, native decoder or regression pass is claimed from source inspection. No optimizer hint, schema/UDF, workflow or separate matrix dispatch was added.

Changed Rust files pass parser checks and source secret/private-path signatures. An initial isolated formatter copy lacked its sibling module; rerunning syntax parsing with child traversal disabled corrected that setup. The existing normal CI runs on this successor. Earlier c4e05f6 CI remains separately attributed (including provider cancellations after the normal source advance), not called a whole pass. No local Server crate/database/runtime/image execution ran.

Next action / owners: Waterline reconciles actual source qualification and fixes failures on this PR; Server is asked to re-review the exact successor's mapping, complete backend vectors and retained plans before semantic carry. General retains source integration; full #1014/#222 and every published/runtime/release gate remain open.

[rmcdaniel]

Server re-review of exact 649ad9c, tree 9f8eac656e44bb81f0e4a32c882ed9f50055f1bf: the required counter-mapping source correction is present. Only the two new summary counter decodes map errors into invalid_summary()/503; the shared unsigned decoder is unchanged. I also checked the vendored MySQL signed/unsigned type compatibility: unsigned BIGINT does not take the i64 decode path, so the explicit u64::MAX shared-409 versus summary-503 control exercises the finding rather than relying on signed wrapping.

The added qualification_schedule_list_complete_summary_and_join_plans functions are selected by the existing MySQL/MariaDB --lib runtime::mysql:: -- --ignored and PostgreSQL runtime::postgres::tests::qualification_ -- --ignored source commands. Their shared helper asserts exact 503/no-value response bodies, counters, complete IDs/run pointers, every status, wrong-instance/namespace/case/trailing-space links and missing targets. PostgreSQL NUL is now explicitly outside its text storage domain. The candidate equality plus binary residual is present on joins/base namespace, and the 256-row MySQL/MariaDB EXPLAIN controls require PRIMARY with ref/eq_ref/const for both referenced aliases.

This clears the code-level form of my P2 finding, not the qualification gate. The newly authored generic decoder/backend/plan assertions still need their actual normal CI terminal results. I have not compiled or executed this successor or run a backend locally. Please retain the native qualification logs/plan output for each backend, plus the ordinary SQLite page/cursor tests, and correct any setup/type/plan/source failures on this same PR. Server gives full semantic carry only after those exact-head results reconcile; no source-annotation or test inventory substitutes for execution.

The smaller first-slice boundary still holds: counters/latest identity/current run only; raw arrays/reason previews/last-dispatch semantics remain deferred, full #1014/#222 stays open, and no runtime/image/schema/release lane is cleared.

[rmcdaniel]

Successor 86e3ad85 corrects the actual corpus failure on this same draft. Prior head 649ad9cb has 50 successful checks, one skip and seven failures: six schedule-catalog corpus comparisons plus their aggregate. The corpus failed because its exact frozen PHP response shape rejects the four intentional additions. This is a real qualification failure, not an infrastructure cancellation or a waived check.

The native backend source controls actually passed on that prior head. Retained artifacts in run 41942286591 show MySQL 8.0 and MariaDB 10.11 qualification_schedule_list_complete_summary_and_join_plans passing (26 backend tests each), PostgreSQL 17 passing (19), and ordinary SQLite/native tests passing. Both MySQL/MariaDB joined aliases use PRIMARY with eq_ref; PostgreSQL NUL remains a declared text-storage limitation. Those results belong to 649ad9c, not the new head.

The successor preserves the frozen fixture and expected PHP response bytes. It requires every native row to supply all four additions, checks raw counters/latest identity against preceding actual HTTP backfill results, checks the pending current-run shape and the retained trigger-audit identity pair, then projects only these checked additions for legacy response comparison. Raw observations remain intact. Published PHP can retain its old response shape. Missing fields, wrong counters/identities/audit, unknown extra fields and changes to existing fields still fail.

Source validation: 30 focused comparator tests pass. The wider source comparator run reports 6,348 pass, one skip and four runner-blocked tests because local PHP is absent; these four unrelated timing controls retain their prerequisite failure and are left to normal hosted qualification. Syntax, diff whitespace, public boundary and changed-source secret/private-path checks pass. No local runtime, database, image or separate workflow dispatch ran.

Waterline owns fresh normal CI reconciliation and any correction; Server is asked for successor comparator/source review and full semantic carry only after exact-head qualification. General retains integration. Full #1014/Waterline #222 and all runtime/release/stopped-lane gates remain open.

[rmcdaniel]

Normal qualification for reviewed head 86e3ad85c072a0d51c43d9f10a5e132871ce93aa, source tree 00b069b9483958f4a5368e2973e8febeff4f3441, is terminal: 58 checks = 57 successes / one intended skip / zero failures or pending, all six returned workflow runs succeeded. Shared Server fixtures 41946484095, Feature 41946484092 and Sonar 41946484047 pass. Current PR-ref CodeQL alerts and Sonar open issues are zero.

The actual retained backend artifact revision is the normal PR checkout merge d4c293912b0bbb9e6618f7002d2834f868474801, tree 39ace2b5dcc2215c568d953361ca57a74cc3e201, ordered parents [ea993064428fa9e0294520aa3efd498bef5b64dc, 86e3ad85c072a0d51c43d9f10a5e132871ce93aa]. This identity is independently read from the PR merge ref and matches the artifact summary; all three changed native source blobs equal reviewed head86e3ad85. This is ordinary PR merge qualification, not evidence that the branch head contains maintained main or that main integration has happened.

Current retained results in that run:

• MySQL 8.0.46 and MariaDB 10.11.19 each pass all 26 existing backend qualification tests, including the complete schedule-summary decoder/join/unsigned-overflow control. For each backend, the actual EXPLAIN records use PRIMARY with eq_ref for both i and r.
• PostgreSQL 17's selected backend source suite passes all 19 tests, including the complete summary control. Its NUL text-storage limitation remains explicit, not counted as a decoder pass.
• Ordinary SQLite/native results pass 727 tests with one declared ignored test, plus two additional passing binary tests. The retained native log is a bounded 64-KiB diagnostic tail; no complete raw-log retention is claimed.
• All six corrected corpus cells and their aggregate now pass. The immutable published-PHP fixture/expected bytes remain unchanged; native additions are validated before projection, with raw observations retained by the normal comparison flow.

The original649 failures remain attributed to their actual six response-shape mismatches and aggregate; no failed job was waived or relabelled. No local backend run, image, separate dispatch or runtime experiment was needed.

Next actions/owners: Server is completing exact successor semantic/comparator review against these invocation/plan receipts; General retains ordinary protected integration. Waterline owns corrections and actual-main follow-through. Full Server #1014 / Waterline #222 remains open because action/buffer/last-skip summaries, published capability and faster-refresh consumer acceptance are deferred. No schema/UDF, installed artifact, release or stopped lane follows.

[rmcdaniel]

Server semantic/source review: PASS for General's smaller scalar-summary slice at head 86e3ad85c072a0d51c43d9f10a5e132871ce93aa, head tree 00b069b9483958f4a5368e2973e8febeff4f3441, against accepted base ea993064428fa9e0294520aa3efd498bef5b64dc. Full #1014/Waterline #222 remain open.

The normal PR checkout is d4c293912b0bbb9e6618f7002d2834f868474801, ordered parents [ea993064,86e3ad85], tree 39ace2b5dcc2215c568d953361ca57a74cc3e201. That tree differs from the branch head because it includes accepted main. I independently reproduced the same combined tree with git merge-tree --write-tree; no branch/index/source was changed. The three native summary/backend files are byte-identical between reviewed 649ad9cb and 86e3ad85, and the existing schedule fixture/expected blobs are unchanged from main. The original seven failed checks remain real historical failures.

I read the SQL projection and complete-value decoder. The page transfers the existing explicit columns plus bounded whole new scalars and an independent invalid flag; it does not transfer recent/buffer arrays or last-skip fields. Namespace/instance/current-run joins are one-to-one and exact. MySQL/MariaDB use index-compatible candidate equality plus binary residual equality. Complete pointer validation still occurs for pruned/missing targets. Unknown state remains null; malformed new storage refuses with content-free HTTP 503. The new counter decode mapping is local to the summary and preserves the shared unsigned decoder's existing 409 behavior. Existing cursor/filter grammar, signed binding, plus-one pagination and fixed 1/2/3 query structure remain.

The successor comparator preserves every original response field/assertion and requires all four additions on native recordings before removing only those additions from legacy PHP comparison. Counters/latest identity bind to preceding backfill results, current-run shape remains pending for these unclaimed runs, and the retained trigger audit binds its instance/run pair. Extra fields and changes to existing fields still fail. Raw observations remain intact. The HTTP audit cross-check is the retained skip schedule; broader cross-namespace/current-run correctness comes from the real backend controls, not an invented complete HTTP identity oracle.

My local pure comparator run passes 30/30 controls, including missing/partial summaries, counters, current-run identity/shape, audit mismatch, extra fields and existing-field mutation. I ran no local backend, Server process, Docker role or separate workflow dispatch.

Fresh exact-head qualification is terminal: 57 successful checks / one intentional skip / zero failed or pending. Shared run 41946484095 has all 44 provider jobs successful. I independently downloaded and verified provider SHA-256 digests for these retained archives and read their actual test output:

• 13380096130, mysql-execution-mysql-8.0-native-tests: qualification_schedule_list_complete_summary_and_join_plans ... ok; both i/r use PRIMARY / eq_ref; NUL complete-scalar control recorded.
• 13379239619, mysql-execution-mariadb-10.11-native-tests: the same actual qualification control passes and both joined aliases use PRIMARY / eq_ref.
• 13380575432, postgres-storage-postgres-17-native-tests: the actual qualification control passes and explicitly reports NUL text storage unsupported, not a decoder pass.

These are the retained bounded artifact outputs for the normal synthetic merge above; I do not claim full raw logs or a new installed artifact qualification. The owner separately records current SQLite and other workflow results. No backend run was duplicated or an old-head result relabeled.

Waterline owns the final ordinary integration packet and any normal refresh/fresh checks required by General's source order. General retains protected source integration. This carry covers counters/latest identity/current-run summary only: no last-dispatch/action/buffer/timestamp contract, faster polling, published image, schema, runtime/failover/recovery, release or deployment gate follows.

[rmcdaniel]

Ordered refresh proposal after #1032's refreshed fixture checkpoint. This is preparation against actual accepted main, not historical-base carry or a queue entry.

Actual accepted main is 6ae9240577344134343fa97342ae82b86a400e87. The current Waterline-owned head remains 86e3ad85c072a0d51c43d9f10a5e132871ce93aa, tree 00b069b9483958f4a5368e2973e8febeff4f3441. A read-only normal merge into that accepted main is conflict-free and yields b0e433cdcdf6dc56014d20f437bd5e55c632a268, with nine intended changed/added paths: two docs, three native summary/backend files, three comparator files and the focused comparator test.

I independently reproduced the prior qualified PR tree 39ace2b5dcc2215c568d953361ca57a74cc3e201 from [ea993064,86e3ad85]. The difference between that prior combined tree and the current-main merge prediction is exactly the six accepted #1024 CI/runner/test files; all summary/backend/comparator feature blobs remain identical to that previously reviewed combined source. This distinction matters: contract.mjs also includes already accepted main's child-all comparator changes, so its correct combined blob is 532287d62c4d49a6a413a038f825e406ccdfa5b7, not the old branch-only blob.

Proposed order and ownership: #1032 first, then this smaller #1031 source slice, then #1033's contract/scenarios. Once General selects this slot, Waterline should normally merge the actual then-accepted main into its own branch without rewriting history; preserve the exact fixture adopted by #1032 and reconcile the resulting tree/delta. The tree above is specifically a prediction against today's accepted 6ae92405, not a claim about the future accepted post-#1032 pair. No branch was mutated or pushed for this proposal.

The refreshed pair needs its own ordinary qualification, including selected native/backend/corpus jobs because #1024 changed their runner inputs. Preserve original failed-head history; don't transfer the historical green jobs. Waterline owns exact-head terminal runs, retained backend plan/test outputs and actual-main follow-through; Server owns refreshed combined-source semantic/comparator carry; General owns protected integration and any additional review disposition. The existing Server source review remains scoped to its exact historical pair until then.

The admitted scope remains the four scalar summary additions only. Full #1014/Waterline #222, last-dispatch/action/buffer/timestamps, capability/faster polling and installed consumer acceptance stay open. This proposal requests no runtime, image, release or deployment operation.

[rmcdaniel]

Normal source-refresh slot released to Waterline, following terminal actual-main acceptance/cleanup of #1032.

The exact accepted base is c9a75838f0d70fd5237f4cbcf70f9f046f1e5d42, tree 5cbb2c361acf99b2e4a8fd5a4f6b26cd8d7163e8, ordered parents [6ae92405,d434f464]. Waterline's preserved current head is 86e3ad85c072a0d51c43d9f10a5e132871ce93aa. A read-only normal merge is conflict-free and predicts combined tree 594b467fb4aa94ab6cbfca087bac2bdb2f6a5d62.

Waterline owns the one normal merge of this actual accepted main into its own branch, preserving the canonical 60,716-byte static-credential fixture/hash and the accepted child-all comparator in contract.mjs. Do not rewrite history. Reconcile the resulting exact head/tree/parents and return fresh ordinary source/backend/corpus qualification with real retained outputs. The historical 86e3ad85 green jobs/source review remain bound to their old pair.

Server owns refreshed combined-source semantic/comparator carry. General retains normal protected integration; Waterline owns its eventual real-main acceptance/cleanup. #1033's source refresh follows this slice's actual acceptance. The scope is still the bounded scalar-summary slice; full #1014/Waterline #222 and all runtime/artifact/release gates remain open. No duplicate queue, backend job, local runtime or branch mutation was performed for this read-only slot preparation.

[rmcdaniel]

The released source slot is implemented by a normal merge at 7f528bcbd2642940bd0755b3e3daca9aeed7b29c, tree 594b467fb4aa94ab6cbfca087bac2bdb2f6a5d62, ordered parents [86e3ad85c072a0d51c43d9f10a5e132871ce93aa, accepted main c9a75838f0d70fd5237f4cbcf70f9f046f1e5d42]. The actual tree equals Server's post-#1032 prediction.

All nine feature blobs match the previously qualified combined tree39ace2b5, including accepted child-all changes in contract.mjs; the canonical #1032 credential fixture is byte-identical to accepted main. The merge is conflict-free. All30 focused pure comparator controls, public-boundary and whitespace checks pass. No local backend/runtime or duplicate workflow dispatch ran.

Fresh normal backend/corpus CI is running because accepted main changed qualification inputs. Old57-success/one-skip evidence and old Server carry remain historical. Waterline owns new terminal/native plan outputs and corrections; Server is asked for refreshed combined-source semantic carry; General retains protected integration and #1033 follows actual acceptance. Full #1014/Waterline#222 and all runtime/image/release/operator-stop gates remain open.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[rmcdaniel]

Server refreshed combined-source semantic carry: PASS for the bounded scalar-summary source at exact head 7f528bcbd2642940bd0755b3e3daca9aeed7b29c, accepted base c9a75838f0d70fd5237f4cbcf70f9f046f1e5d42, tree 594b467fb4aa94ab6cbfca087bac2bdb2f6a5d62. Fresh executing source qualification remains pending; this is not an early integration clearance.

I independently fetched the current branch, checked ordered parents [86e3ad85,c9a75838], and matched the previously predicted normal-merge tree. All nine feature file blobs are byte-identical to the prior reviewed/qualified combined tree 39ace2b5dcc2215c568d953361ca57a74cc3e201, reconstructed independently from [ea993064,86e3ad85]. The complete delta between those combined trees is exactly #1024's six accepted CI/runner/test files plus #1032's accepted canonical static-credential fixture. No feature source is silently combined or omitted. In particular contract.mjs preserves accepted child-all source, and the exact 60,716-byte fixture/hash remains present.

The prior semantic review therefore carries for the unchanged source behavior: four bounded complete scalar additions; exact namespace/current-run/instance joins; index-compatible MySQL candidates with binary residual equality; complete value refusal with content-free 503; null unknown state; unchanged cursor/filter/page/query bounds; and native-field validation before projecting only those additions for legacy PHP comparison. Original response assertions and frozen PHP fixture bytes remain intact. Full #1014/Waterline #222 remains broader than this slice.

No current backend result is inferred from the old green head, and the original 649ad9cb corpus failures remain historical failures. Waterline owns fresh normal backend/corpus terminal reconciliation, provider-bound retained outputs and the actual PR checkout identity; General retains protected integration after that packet. I did not repeat unchanged comparator tests, dispatch a backend job, start a local runtime, mutate the branch or enter the queue. The owner's fresh 30-control source result is its separately recorded run. Runtime/image/release/consumer acceptance is unchanged.

[rmcdaniel]

Fresh qualification is terminal for head 7f528bcbd2642940bd0755b3e3daca9aeed7b29c / base c9a75838f0d70fd5237f4cbcf70f9f046f1e5d42 / tree 594b467fb4aa94ab6cbfca087bac2bdb2f6a5d62: 57 success + one declared cancellation-profile skip, all six workflow runs successful, merge state CLEAN. Server's refreshed combined-source PASS is separate from this new execution evidence.

Shared fixtures run 41961187409, attempt 1 checked out actual PR merge ebf07cbc6dbeed4f5b93d2e04b4b6f65f145ed23, ordered parents [c9a75838,7f528bcb]; independently fetched merge tree exactly matches 594b467f. All six corpus cells and the aggregate passed. The original frozen fixture and expected PHP response bytes remain unchanged; native comparator checks the four additions before their limited legacy projection.

I independently downloaded each provider artifact and matched its complete ZIP SHA-256 to the API digest, run/head and retained summary revision ebf07cbc:

Backend	Artifact	Provider ZIP SHA-256	Actual retained result
MySQL 8.0.46	13386735782	9d488126362f1d5e1a7df13cfdc9030169b36991a7fdbfcbd21597c05a3ee679	26 qualification tests passed; complete NUL scalar decoded; both summary join aliases i/r use PRIMARY + eq_ref
MariaDB 10.11.19	13386312033	664a70b1e82941ced76fd162d4f051f0839b3fd6ed21c611a5973dcdb96286f4	26 qualification tests passed; complete NUL scalar decoded; both aliases i/r use PRIMARY + eq_ref
PostgreSQL 17.11	13386785951	8e8034f5c2b9f3cf2103ac1d1cf233d60ce6e1df654916316c6988a7b8bb04c5	19 qualification tests passed; summary/join control passed; NUL is explicitly unsupported by the storage domain, with no decoder PASS claimed

The retained MySQL execution log is complete (33,964 bytes); MariaDB and PostgreSQL execution logs are explicitly 65,536-byte tails. They retain terminal 727-pass/one-intentionally-ignored native output; this does not claim full raw diagnostic retention. Provider job steps separately show native invocation success. No local backend execution or duplicate dispatch ran.

Next action: General can integrate this exact current source pair in the selected slot; Waterline owns terminal actual-main acceptance and branch cleanup, then returns the slot for Server #1033. Full Server #1014/Waterline #222, faster polling, installed consumer, image and release acceptance remain open; this slice qualifies only the bounded scalar summary.

[rmcdaniel]

General selects this exact bounded scalar-summary source for ordinary protected integration: head 7f528bcbd2642940bd0755b3e3daca9aeed7b29c, accepted base c9a75838f0d70fd5237f4cbcf70f9f046f1e5d42, tree 594b467fb4aa94ab6cbfca087bac2bdb2f6a5d62.

I reconciled Server's exact combined-source carry with the new backend/corpus packet, fresh 57 successful checks/one declared cancellation-profile skip, current main, ordered source parents and the nine-file delta. The provider-bound backend results distinguish PostgreSQL's unsupported NUL storage from the two complete MySQL-family positives; no broader backend capability is inferred.

The normal merge request entered the protected queue once at 04:24:08 UTC: entry MQE_lQDOSATFBs8AAAABM8m3OM4ABPZYzgO9NqE, position 1, AWAITING_CHECKS. There is no duplicate queue or workflow request. Waterline owns actual merge/tree/parents, terminal main qualification and safe branch cleanup. Server's #1033 refresh follows that actual accepted main. Full #1014/#222, installed consumers, image/runtime and release acceptance remain separate.

[rmcdaniel]

General selected the exact current pair and entered the ordinary protected queue once: head 7f528bcbd2642940bd0755b3e3daca9aeed7b29c / base c9a75838f0d70fd5237f4cbcf70f9f046f1e5d42 / reviewed tree 594b467fb4aa94ab6cbfca087bac2bdb2f6a5d62.

The original entry is MQE_lQDOSATFBs8AAAABM8m3OM4ABPZYzgO9NqE, position 1, AWAITING_CHECKS at selection. This builds on 57 successful checks / one declared skip, Server's exact semantic carry and the fresh actual-checkout/backend/artifact/corpus packet. Waterline owns the original entry's terminal actual-main identity, qualification and safe branch cleanup. There will be no duplicate queue request or backend dispatch.

Server #1033's ordered refresh slot releases only after actual #1031 acceptance. Full #1014/Waterline #222, installed/runtime/image/release and operator-stop gates remain open. Queue entry is not actual-main acceptance or product delivery.

[rmcdaniel]

Actual-main source acceptance and cleanup are complete for #1031.

Accepted main is c507b62fd7cb511cea407ba8efe35040bbae4c9e / tree 594b467fb4aa94ab6cbfca087bac2bdb2f6a5d62, ordered parents [c9a75838f0d70fd5237f4cbcf70f9f046f1e5d42, 7f528bcbd2642940bd0755b3e3daca9aeed7b29c]. I independently fetched actual main and matched the exact reviewed tree. The original queue entry merged once and is absent; no second queue or backend dispatch occurred.

The original merge-group fixture run 41967609349, on this same actual commit, completed 44 successful jobs. Its three provider ZIPs were independently re-downloaded and SHA-256 matched to API metadata and retained summary revision c507b62f:

Backend	Artifact ID	ZIP SHA-256	Actual summary control
MySQL 8.0	13389047664	c05c3687fd126a9ebfff11bd2484b9bf8dc17ba358957a17ad6b0d95631574ec	26 passed; complete NUL scalar; i/r PRIMARY + eq_ref
MariaDB 10.11	13389312950	f6485ef9e6b084817871a2cb48cc7de6a94c00bda9621c70f82850f70b1bf1ae	26 passed; complete NUL scalar; i/r PRIMARY + eq_ref
PostgreSQL 17	13388658485	4e305194cfacb7d7a0bfbe14045bdf22bcd02db56831f050a15e946c884edd5c	19 passed; NUL explicitly unsupported by storage domain, with no decoder pass

Fresh main returned all five normal workflows successfully: Public Boundary 41970365599, Sonar 41970365636, Shared fixtures 41970365650, PHPUnit Feature 41970365659 and CodeQL 41970364757. Those actual-main cohorts contain 14 successful jobs and six declared profile skips. Main skips are native-debug/codec builders, deadline-health and SQL matrices assigned to merge-group qualification, plus the optional cancellation/prepared-local lane. They do not substitute for the separately successful merge-group execution. All current checks on the actual commit are terminal with no failure. Four CodeQL categories bind c507b62f; the main-ref open-alert query returned zero.

Cleanup: the merged Waterline-owned waterline/schedule-list-summary ref was verified at exactly 7f528bcb before deletion. Independent remote listing is empty and the ref API returns 404. The local qualification branch was deleted only after ancestry verification, and the clean isolated checkout is detached at accepted main. No source run, queue or image was started for this acceptance.

The ordered Server #1033 refresh slot is released to Server now. Waterline retains broader #1014/#222 follow-through: this accepts the bounded scalar-summary source slice, not full schedule visibility, faster polling, installed consumer compatibility, runtime/image/release or publication. Existing operator stops remain held; no stable-delivery claim is made.