Internal maintenance — no functional change. The recent performance work added some temporary timing measurements to the system-status response so that performance captures could pinpoint exactly which part of a slow page was responsible. Now that the performance work is complete, those measurements have been removed to keep the status response clean. This has no visible effect: the timing fields were never displayed anywhere in the interface; they only rode along in the raw response for diagnosis. A leftover internal time-unit inconsistency in one status calculation was also tidied up at the same time — it produced the correct result already, so nothing changes for users.
Behind the scenes
- Removed the per-section endpoint checkpoints (
_endpoint_timings_ms), the per-query stats timings (stats_timings_ms), and the capacity-probe timings (_timings_ms/_outer_call_ms) from the/api/statusresponse and the capacity module. All of the underlying queries are unchanged; only the surrounding stopwatch code was removed. Confirmed nothing in the frontend read these fields. - Corrected the one remaining place where an hour-bucket window was divided by 3600 before comparison (
all_totalin the status path). Because the hourly rollup is already pruned to the retention window, the old and new forms select the same rows, so the displayed total is identical — this is purely removing the last instance of the units inconsistency so none remains in the tree.
Operational notes
- No schema changes, no config changes, no migration. No user-visible behavior change.