Replacement stack for PR #21192; Linux headless integrated service proof.
Lower head88465a1d8935ec501cbc04bd680ee98e993d4eb5;upperba4991b0ac1ea54a22bcf99eccda90eaec9ce76e;actualincludedbasea1e39174493372daca48e6d59545a7681e64d0f6.
Fresh Astra tests47→59 across2existingfiles:cadence3→13,reader44→46,12cases3tabletests. RED recent/futurestartup againstmain andevents/receiptsreader routing againstremovedreadOnly;GREENall59.
Fresh Sol implementation lowerstraightline completionpersistence,upperexistingreadOnly extensionpoint. Scopedservertypecheckpass zeroerrors/warnings,154suggestions;lint/fmtpass.
Primary integrated run:
note: You are running `vp test` as a Vite+ built-in command. If you meant to run the test npm script, use `vpr test` instead.

 RUN  v5.0.1 /home/ubuntu/wt/usage-billing-01


 Test Files  2 passed (2)
      Tests  59 passed (59)
   Start at  19:05:37
   Duration  18.04s (tests 42%, transform 40%, import 18%, setup 1%)

  Transform  transforming modules took 8.64s · 40% of tracked time, re-done on every run
             persist transforms across runs with fsModuleCache: true
             learn more: https://vitest.dev/guide/improving-performance#caching-between-reruns


Real temporary timestamp I/O + actual file-backed SQLite reader-worker concurrency; controlledEffectclock/Deferred/queues, no polling/sleeps. Successfulcompletiontimestamp includingzero-deletes;failedcompactiondoesnotadvance;writefailurebenign;restartdelaybounded1min..6h;heldreaderblocksdiscoverywhileunrelatedwritercommits;releaseallowslegacydeleteandpendingretention. In-memoryfallbackexistingtestsretained.
No live full server, providers or client run for this backend maintenance change. No CPU, memory, event-loop-latency benchmark claimed; original macOS measurements not reproduced. Native macOS/Windows/Electron/iOS/Android unverified.

Follow-up at upper ba4991b0ac1ea54a22bcf99eccda90eaec9ce76e: CodeRabbit correctly identified that the earlier fork/clock hop did not explicitly checkpoint discovery. Fresh Astra strengthened the existing two cases without adding cases or production hooks. A scoped Effect SQL transformer observes the discovery query attempt before acquisition, then the real adapter span proves the acquired connection is the reader. An unrelated writer insert completes while the reader transaction is held.
Removing both readOnly applications fails both cases at expected reader / actual writer (not timeout), even though the earlier negative completion assertion passes. Production restored exactly; 59/59 integrated tests, lint, format, and scoped server typecheck pass. Fresh implementation agent confirmed the mechanism and reran both cases successfully.

Final integrated green output:
note: You are running `vp test` as a Vite+ built-in command. If you meant to run the test npm script, use `vpr test` instead.

 RUN  v5.0.1 /home/ubuntu/wt/usage-billing-01


 Test Files  2 passed (2)
      Tests  59 passed (59)
   Start at  19:29:02
   Duration  21.26s (transform 45%, tests 37%, import 17%, setup 1%)

  Transform  transforming modules took 12.09s · 45% of tracked time, re-done on every run
             persist transforms across runs with fsModuleCache: true
             learn more: https://vitest.dev/guide/improving-performance#caching-between-reruns