ERD_INDEXERS.md defines job cadence expectations for derived-table refresh workflows and expects deterministic refresh timing.
Existing tests validated specific refresh behavior (connectivity/reputation/base-score/canonical backfill), but did not explicitly validate seeded job_schedule.cadence_seconds against ERD timings.
Phase 9 still had Ensure derived tables refresh according to ERD timing and caching rules unchecked.
Decision:
Add a data-layer test in revaer-data to assert job_schedule cadence values for all indexer jobs that drive derived refresh and related maintenance windows.
Keep coverage dependency-free and proc-centric:
job_schedule_cadence_matches_erd_refresh_timing validates configured cadence seconds for refresh, rollup, GC, purge, and RSS schedules.
Alternatives considered:
infer timing correctness from runtime behavior only: rejected because explicit cadence drift can pass behavioral tests but violate ERD schedule requirements.
Consequences:
Positive outcomes:
ERD timing expectations are now executable and regression-safe;
derived refresh cadence drift will fail tests early.
Risks or trade-offs:
if ERD cadence values change, this test and migration seeds must be updated in the same change.
Follow-up:
Implementation tasks:
Continue with the next unchecked acceptance/hard-blocker item.