Add a focused Linux InfluxDB readback qualification mode

Expose the existing real million-record InfluxDB native case and terminal diagnostic through manual mode=influxdb-readback with a five-minute native deadline. Keep the full eight-native-control gate for normal push/full/contracts runs and keep focused results separate from full-session qualification and public figures.

Validation: actionlint, repository governance and git diff --check. C# workload and adapter bytes are unchanged from the verified readback repair.

Author: KSemenenko

.github/workflows/benchmarks.yml

@@ -5,9 +5,9 @@ on:
   workflow_dispatch:
     inputs:
       mode:
-        description: Run harness/native controls or the full database suite
+        description: Run native controls, focused InfluxDB readback, or the full database suite
         type: choice
-        options: [contracts, full]
+        options: [contracts, influxdb-readback, full]
         default: full
 permissions: {contents: read, actions: read}
 concurrency:
@@ -110,7 +110,11 @@ jobs:
           compression-level: 0
           retention-days: 90
       - name: Test external database bulk operations and joined cleanup
+        if: inputs.mode != 'influxdb-readback'
         run: dotnet benchmarks/KeyLoad.ComparisonHost/bin/Release/net10.0/KeyLoad.ComparisonHost.dll --output Detailed --maximum-parallel-tests 1 --minimum-expected-tests 8 --treenode-filter '/*/*/NativeDatabaseTests/*' --results-directory "$GITHUB_WORKSPACE/benchmarks/KeyLoad.ComparisonHost/bin/Release/net10.0/TestResults/NativeAdapters" --report-trx --timeout 40m
+      - name: Test InfluxDB million-record readback and joined cleanup
+        if: github.event_name == 'workflow_dispatch' && inputs.mode == 'influxdb-readback'
+        run: dotnet benchmarks/KeyLoad.ComparisonHost/bin/Release/net10.0/KeyLoad.ComparisonHost.dll --output Detailed --maximum-parallel-tests 1 --minimum-expected-tests 1 --treenode-filter '/*/*/NativeDatabaseTests/InfluxBulkReadbackPreservesCompleteBodiesAndHealthyContinuation' --results-directory "$GITHUB_WORKSPACE/benchmarks/KeyLoad.ComparisonHost/bin/Release/net10.0/TestResults/NativeAdapters" --report-trx --timeout 5m
       - name: Test bounded terminal execution diagnostics
         run: dotnet benchmarks/KeyLoad.ComparisonHost/bin/Release/net10.0/KeyLoad.ComparisonHost.dll --output Detailed --minimum-expected-tests 1 --treenode-filter '/*/*/ComparisonHostCaptureTests/*' --results-directory "$GITHUB_WORKSPACE/benchmarks/KeyLoad.ComparisonHost/bin/Release/net10.0/TestResults/TerminalDiagnostics" --report-trx
       - name: Save original preparation diagnostics

docs/Features/BenchmarkComparisons/SessionExecution.md

@@ -167,8 +167,14 @@ matrix, restarts or separate benchmark scenario.
   Manual `mode=contracts` runs the same native and harness controls, retaining test
   diagnostics. Workload fan-out, session measurements
   and Website dispatch belong to `mode=full` and push runs.
+  Manual `mode=influxdb-readback` runs only the existing InfluxDB million-record
+  native case and terminal diagnostic control, with a five-minute bound on that
+  native case. It retains original Linux resource and test evidence without the
+  other databases, workload fan-out or public figures. This focused check does not
+  qualify the full eight-control gate or complete session suite; those gates stay
+  unchanged for `mode=contracts`, `mode=full` and push runs.
   A newer run of the same branch cancels the superseded run instead of waiting
-  behind its measurements, including when changing between contract-only and full runs.
+  behind its measurements, including when changing between manual modes and full runs.
 - Ingestion at 1/10/500 clients, the minimum measured operation counts, recall,
   replicated acknowledgement, resource bounds and all required workload families are
   kept. The number of jobs is not the number of measured scenarios.