Phase 7.2 — Emergency hardware stabilisation

Three hardware faults corrected while the machines are idle.

An undersized battery backup swapped for a larger one — one server was drawing 103% of its unit's rating. A failed write-cache capacitor replaced. And a disk controller setting that is causing measurable stalls today switched off, with the controller's idle 2 GB cache enabled in its place.

All three need the chassis opened or powered down, which is exactly why they happen now.

Downtime: chassis power-offs only — and everything is already off.