Disk usage climbed to 97% from unbounded temp and log growth during the nightly ETL job. Real, unrelated changes exist elsewhere in the tenant during this window — the platform doesn't blame any of them.
Steps to test
- Open INC0000009.
- Steps 4–7: node_disk_usage_percent ramps to 97% over the final two hours.
- Step 8: explicitly notes no change on this CI was found, despite unrelated changes existing elsewhere.
- Steps 13–14: fixed live, verification outcome PARTIALLY_RESOLVED — disk recovered, but the alert prevention step wasn't added.
Outcome: the platform actively declines to force-fit an unrelated change as the cause.