v0.9.54-alpha — Guard Dog catches slow-disk stalls + CloudTAK update reliability

Guard Dog now catches the slow-disk problem that silently takes TAK stacks down

On shared and cloud hosting, a disk can pass an ordinary speed test while individual database writes stall for seconds at a time. That small-write stall — not a throughput drop — is what quietly snowballs into an Authentik/login slowdown. Until now Guard Dog only watched bulk throughput, so this failure mode could run with no warning.

What's new:

• Commit-latency monitoring. Guard Dog now measures small-write (4 KB) commit latency directly — the metric that actually governs database responsiveness — charts it alongside throughput, and warns you when it crosses a safe ceiling.
• "Provider disk contention — act now" alert. When slow commits coincide with the database having clients waiting, Guard Dog sends a distinct early-warning alert that attributes the cause to host disk contention and points you at your provider before an outage instead of after.
• Reports that match your provider's numbers. The disk report now includes the small-write latency figures, so what Guard Dog shows lines up with the numbers a hosting provider quotes.

The existing throughput benchmark and its alerts are unchanged — this is added alongside them.

CloudTAK "Update Now" reliability

CloudTAK's upstream sometimes ships a release whose internal version label trails the release tag, which made the console believe an update never took — leaving a stuck "update available" badge even on a fully up-to-date box. The console now reads the release correctly: the badge clears after a successful update and the true version is shown.

Upgrade

Applied automatically. The new disk-latency watch rides Guard Dog's existing schedule; the CloudTAK fix takes effect on the next console restart. No configuration required.