TODO grooming — close the RS-11.6/RS-12 arcs, add next-session sequencing

[dorkmo]

TODO grooming: closes the RS-11.6/RS-12 arcs as documented entries, adds RS-12.9 next-session sequencing, and (2026-08-22) records the session outcomes: RS-3.3 camera legs, RS-11.8 chantab findings, PM-1 tracking, 927.5 clean 4/4.

MERGE ORDER (from review): merge #111 and #110 BEFORE this PR. This TODO references the POWER_MANAGEMENT.md design (lands with #110) and the session evidence directories, seq= logging change, and CI heal (land with #111). Merging #109 first would publish dangling references; after those two land, this PR applies cleanly with every reference resolving.

Review status: all five Copilot comments addressed -- the RS-12 banner now names the full env-gated mitigation (flag + strict-hold gap), and the ordering dependency is declared here and in the banner text itself.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Documentation-only PR that performs end-of-session TODO grooming after the RS-11.6/RS-12 investigation work (per #108), updating campaign status, sequencing notes for the next bench session, and superseding older TODO entries with pointers.

Changes:

• Adds a dated “Radio status” banner to the root LifeTrac-v25/TODO.md summarizing current bench loss-floor state and next steps.
• Updates DESIGN-CONTROLLER/TODO.md to mark RS-12 search phase as closed, add RS-12.9 next-session sequencing, and annotate RS-3.6 status plus RS-11.6 superseded items.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
LifeTrac-v25/TODO.md	Adds a top-of-file radio-status banner summarizing current campaign state and next actions.
LifeTrac-v25/DESIGN-CONTROLLER/TODO.md	Closes RS-12 search arc, adds next-session sequencing, and updates related RS-3.6/RS-11.6 TODO entries.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Suppressed comments (6)

Previously missed (5) — in code that hasn't changed since the last review.

LifeTrac-v25/TODO.md:4

• The 0.9% status is not supported by the cited campaign evidence. The final strict-hold result reports 1.5% loss (also reflected in DESIGN-CONTROLLER/TODO.md:2290), while the n=3 plain-hold result is 1.8%. Publishing 0.9% in the root status banner makes the current operating floor inaccurate.

> The bench LoRa link runs at **0.9 % fragment loss** (from 5.9 % two days

LifeTrac-v25/TODO.md:13

• This calls the next run a “first flight,” but RS-3.3 is already marked complete and its evidence records 472 real-camera frames delivered over LoRa (DESIGN-CONTROLLER/TODO.md:865-877). Describe this as a regression/revalidation run so the root status does not reopen completed work.

> RS-3.6 gate) and the RS-3.3 camera first flight. Live campaign state:

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:918

• This reopens a gate that the committed flash record explicitly marks cleared: bench-evidence/RS_firmware_patch_flash_2026-07-25/NOTES.md:20-25 says the RS-3.6 gate cleared on both boards with REG 0x42 = 0x12. If a required TX/RX transcript was actually omitted, reconcile that evidence record; otherwise close RS-3.6 here rather than queuing a duplicate gate.

  (thousands of TX/RX cycles, tx_fifo_rb_bad=0 throughout). Only the
  FORMAL gate transcript remains (REG 0x42 readback + one TX/RX cycle,
  ~2 min) — queued in the RS-12.9 flash session.

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2320

• This sequencing item contradicts the existing RS-3.3 record: bench-evidence/RS_3_3_real_camera_2026-07-30/RESULTS.md:1-26 says the real camera, encode-to-fit packer, carry fix, age escalation, and liveness valve already ran on hardware and delivered 472 frames over LoRa. Keep the useful lower-loss rerun, but present it as regression coverage rather than an unperformed first flight.

2. **RS-3.3 camera first flight** — `-TxFeed camera` at the current best
   operating point (`-ForceFrfHz <today's pick> -NoParkLast 1`). The
   encode-to-fit packer, carry fix, age-escalation and liveness valve
   have NEVER been on air; flying them on a 0.9%-loss link attributes any
   failure to the path, not the link.

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2263

• The documented floor is 1.5%, not 0.9%. The cited strict-hold evidence and the preserved RS-12 entry at line 2290 both report 1.5%, so this closure headline contradicts its own evidence.

This issue also appears on line 2270 of the same file.

- [x] **RS-12 CLOSED (search phase) 2026-08-17 — mechanism found, host fix
  validated, bench floor 5.9% → 0.9%.** The penultimate-fragment loss was

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2271

• These results do not match RS_12_noparklast_ab_2026-08-17/RESULTS.md: n=3 plain hold measured 1.8% loss/11.6% penultimate share, while strict hold was only n=1 and measured 1.5% loss, 32 timeouts, and 143 published frames. The 0.9%/22/154 figures have no cited evidence, so the closure record should retain the measured arm sizes and values.

  n=3 interleaved + strict-hold legs: penultimate → uniform, loss 0.9%,
  timeouts 22, published 154 (all campaign bests); cost ~−6% offered

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Suppressed comments (4)

Previously missed (3) — in code that hasn't changed since the last review.

LifeTrac-v25/TODO.md:33

• The corrected RS-3.3 evidence in PR #111 says the injector was configured for 24, but only 11 requests landed within the leg (22 received copies). Reporting 24 here overstates the command-plane exercise; update this count to 11.

> leg), 0.3 %, 0.2 % with 24 injected keyframe requests dispatched

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2274

• PR #111's corrected RESULTS.md records 11 injected requests during the leg, received as 22 copies; 24 was only the injector target. This count should be 11 so the closed-entry summary matches its evidence.

This issue also appears on line 2315 of the same file.

  **Update 2026-08-22 (RS-3.3 leg 3): first live-traffic exercise — 24

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2388

• These ranges do not match PR #111's results: the survey recorded 1–7 hits, with maxima mostly −44…−58 dBm and three much louder outliers (−30/−35/−38 dBm). Preserve those values so this TODO does not hide the outliers or misstate the minimum count.

  PR #111): ticker-dominated — 49/50 hot (2–7 hits, flat −46..−52 dBm)
  because the band-wide ~7 s ticker lands in every 30 s dwell;

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2315

• The dependency's corrected evidence shows that only 11 of the configured 24 requests occurred before teardown. Use 11 here; otherwise the session outcome overstates the exercised traffic.

(0.0% / 0.3% / 0.2% — kf-off, kf-idle, kf + 24 injected REQ_KEYFRAME

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

Suppressed comments (3)

Previously missed (3) — in code that hasn't changed since the last review.

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2273

• These figures conflate two experiments. The evidence reports n=3 for the plain hold (1.8% loss, about 40 timeouts, about 144 published frames) and n=1 for the strict 80 ms hold (1.5%, 32, 143); it contains no n=3 strict-hold result at 0.9%/22/154. Please record the two validation levels separately.

- [x] **RS-12 CLOSED (search phase) 2026-08-17 — mechanism found, host fix
  validated, bench floor 5.9% → 0.9%.** The penultimate-fragment loss was

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2289

• This historical opening entry remains unchecked even though the preceding entry closes the RS-12 search phase and the remaining firmware work is separately tracked above and in RS-12.9. That leaves the arc open in rendered checklist summaries, contrary to this PR’s stated purpose; mark the superseded original complete as done for the other historical RS entries below.

- [ ] **RS-12 (original opening entry) — the clustered bulk loss floor

LifeTrac-v25/TODO.md:13

• The cited RS-12 evidence does not support either 0.9% loss or n=3 validation of the strict hold. RS_12_noparklast_ab_2026-08-17/RESULTS.md records three plain NoParkLast fix legs at 1.8%, followed by one strict-hold leg at 1.5%. This banner currently overstates both the measured result and its replication.

> The bench LoRa link runs at **0.9 % fragment loss** (from 5.9 % two days

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Suppressed comments (6)

Previously missed (3) — in code that hasn't changed since the last review.

LifeTrac-v25/TODO.md:4

• The documented RS-12 results do not support a 0.9% floor: the n=3 plain-hold arm averaged 1.8%, and the single strict-hold leg measured 1.5% (RS_12_noparklast_ab_2026-08-17/RESULTS.md:65-71). Use the measured strict-hold value here, or cite the separate archived run that produced 0.9%.

This issue also appears on line 11 of the same file.

> The bench LoRa link runs at **0.9 % fragment loss** (from 5.9 % two days

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2263

• The linked RS-12 evidence reports 1.8% for the n=3 plain-hold arm and 1.5% for the sole strict-hold leg, not 0.9%. Recording 0.9% as the closed campaign floor makes this status disagree with its evidence.

This issue also appears in the following locations of the same file:

• line 2269
• line 2351

  validated, bench floor 5.9% → 0.9%.** The penultimate-fragment loss was

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2291

• The superseded RS-12 opening entry remains an unchecked task, so TODO consumers still see RS-12 as open despite the new CLOSED entry. Mark it complete as done for the other superseded originals in this file (for example, lines 2481, 2527, and 2545).

- [ ] **RS-12 (original opening entry) — the clustered bulk loss floor

LifeTrac-v25/TODO.md:12

• n=3 applies to the plain NoParkLast A/B arm, not to the full flag-plus-80 ms strict hold: the evidence records only one strict-hold leg. This currently overstates how many times the mitigation described on the next line was replicated.

> mitigation validated n=3: `-NoParkLast 1` **plus** the strict hold
> `LIFETRAC_NO_PARK_LAST_GAP_MS=80` — env-gated, NOT a default (see the

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2271

• These strict-hold metrics do not match the cited result: its table records one strict-hold leg at 1.5% loss, 32 timeouts, and 143 published frames, while n=3 describes the earlier plain-hold arm. Please keep the closure summary aligned with the archived measurements.

  event body so commands keep flowing — PR #108 review catch). Confirmed
  n=3 interleaved + strict-hold legs: penultimate → uniform, loss 0.9%,
  timeouts 22, published 154 (all campaign bests); cost ~−6% offered

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:2351

• This historical preflight rationale repeats the unsupported 0.9% measurement. The strict-hold result available at shutdown was 1.5%, so the sequencing record should use that value consistently.

   have NEVER been on air; flying them on a 0.9%-loss link attributes any

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

LifeTrac-v25/TODO.md:8

• The banner overstates the second emitter as characterized. PR #111's RS-11.8 evidence says the −30 dBm 923.5 MHz dwell followed the same ~7 s grid, with only possible interleaved-ticker hints; its summary explicitly says the exact-10 s device B was not directly confirmed. Preserve that limitation so this status does not turn the ~7 s observation into a confirmed second-device fingerprint.

> since: 0.9–1.6 %; from 5.9 % two days prior). RS-11.6: two external ISM emitters characterized (a ~7 s hopper
> and an exact-10 s device at −30 dBm); escaped by carrier choice — the

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

LifeTrac-v25/DESIGN-CONTROLLER/TODO.md:918

• This requeues a formal gate that is already documented as cleared. bench-evidence/RS_firmware_patch_flash_2026-07-25/NOTES.md:10-25 records REG 0x42 = 0x12 on both boards, both Stage 1 gates passing, Method G passing, and explicitly lists RS-3.6 under “Gates cleared.” Keeping RS-3.6 open makes the next-session plan schedule redundant bench work.

  **Status 2026-08-17: FLASHED and field-soaked** — the DIV8 build is what
  both boards have run through the entire RS-11.5→RS-12 campaign
  (thousands of TX/RX cycles, tx_fifo_rb_bad=0 throughout). Only the
  FORMAL gate transcript remains (REG 0x42 readback + one TX/RX cycle,
  ~2 min) — queued in the RS-12.9 flash session.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

LifeTrac-v25/DESIGN-CONTROLLER/bench-evidence/RS_12_noparklast_ab_2026-08-17/RESULTS.md:95

• The archive's params.txt records git_sha=dde2c8a7, but that commit still contains the command-swallowing discard-poll; the fall-through fix lands in the later 5cc4dd9b commit that also adds this archive. Therefore SHA dde2c8a7 does not substantiate the claim that this was the fixed build. Please archive the tested dirty-tree diff/hash (if available) or explicitly qualify the build identity instead of presenting the pre-fix SHA as the tested revision.

normal event body. **Leg L verified the fixed build** (archive
`radio_monitor_20260817_195107_dde2c8a7`, SHA dde2c8a7, bracketed,