Write caching stayed switched off on the repaired array

Replacing the write-cache module restored the cache at the controller, but one of the two disk arrays kept its own caching switched off — and it was the array holding a copy of the cluster database.

The controller reported healthy the whole time, so this looked like a clean repair. It was not: the array that caused the original fault was still running with no write cache, and had begun stalling again. Caching has been re-enabled and verifying it is now part of the procedure, because the top-level health check reads as success either way.