09-Jun-2026 — stop hot paths pulling large uncached payloads #201

build 676 — commit f52b56e

Two independent slow reads were dragging full, oversized documents over throttled clusters on every request.

S3 backup listing (/api/mongo-backup/s3/backups):

• manifestByTimestamp reloaded the entire environments config doc and read the full ~806KB manifest document for every one of 346 backups, just to get each manifest's status field (~272MB transferred to render a dropdown).
• Add completedManifestStatusByTimestamp: one find per site, projected to {timestamp,status} only, run per-environment in parallel and reusing the already-loaded backup config.
• Result: multi-minute hang to ~0.4s for 659 backups.

System config (config?key=system, system-status):

• value.area.groups[].customGeometry (ONS boundary polygons) is 469KB of the 483KB system doc and was read uncached on every page, though only the admin Area Map editor uses it.
• handleQuery now projects customGeometry out of the system read; system-status uses a new systemConfigWithoutGeometry(). Internal consumers (geojson areas) keep the full doc via systemConfig().
• createOrUpdate preserves customGeometry on system saves (matched by groupCode, reorder-safe) so settings saves cannot wipe boundaries; geometry is only changed via the api/areas/* endpoints.
• Area Map editor loads groups from api/areas/groups (geometry intact) instead of the now-stripped config payload.
• Result: system read 483KB/4.3s to 14KB/0.24s; fast on any cluster.