Accelerate Connection Summary Projection
tasks20/20
1. Runtime Projection
- Add connection-aware run summary matching for browser profile keys and explicit connection ids.
- Surface browser-surface failure reason evidence for
surface_failedsummaries. - Reuse connector run pages per connector/status within one summary projection.
- Add a short-lived Postgres single-flight cache for repeated full-list reads.
- Preload retained-size stream/connection projection rows once for the full-list path.
- Add a committed real-browser performance harness with RSC fetch timing and RS/API probes.
- Keep the full connector-summary list shallow for run summaries while retaining deep run summaries on scoped connection/detail reads.
- Add the Postgres source/run spine summary index to bootstrap/migration code.
2. Tests
- Add a sibling browser-run regression test for
listConnectorSummaries(). - Run targeted connection-summary projection tests.
- Run affected spine/control-plane tests.
- Run reference and console type checks.
- Smoke the browser harness against a known-fast live route without true failed requests.
3. Live Verification
- Deploy in a live-stack window.
- Verify
/_ref/connectorsrepeated-call latency before/after. - Verify
/dashboard/runsreal-browser RSC fetch count/timing before/after. - Verify
/dashboardand/dashboard/runscold browser document time after retained-size preloading. - Verify
/dashboardand/dashboard/runscold browser document time after shallow-list projection. - Verify Amazon/Chase row state no longer reflects sibling draft-shell run evidence.
- Verify scoped connection detail/diagnostics still exposes deep run evidence after shallow-list projection.