fix(android): harden startup paths that turned a slow datastore into a hang

[TimeToBuildBob]

Follow-ups from #261. The root cause (quadratic legacy-bucket merge wedging the datastore worker) is fixed in ActivityWatch/aw-server-rust#679; this PR hardens the app-side paths that turned a busy datastore into a blank screen and ANRs, so the next slow datastore operation degrades gracefully instead. Not required for v0.14.1, but cheap and independent.

• BackgroundService: queue the hostname/legacy-bucket migrations once per process. In the emulator repro onStartCommand ran twice on one launch, so the blocking migration command was queued twice on the single datastore worker.
• WebUIFragment: the onReceivedError retry did Thread.sleep(100) on the main thread and reloaded immediately, which turned a slow server start into a tight reload loop on the UI thread. Now a Handler retry with backoff (250 ms → 5 s), reset on a successful page load and cancelled in onDestroyView. Unit test for the backoff.
• WebWatcher: bucket creation moved off the accessibility service's main thread (same pattern MediaWatcher already uses). Play vitals list "Executing service WebWatcher" ANRs parked in Datastore::get_buckets.

Still open after this (separate): SessionEventWatcher/UsageStatsWatcher construct RustInterface on whatever thread creates them, and the JNI getBuckets path unwrap()s when the worker is gone (noted on the aw-server-rust PR).

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; the prior WebView retry finding is resolved and no new actionable defects remain.

Summary

• Queues startup migrations only once per service process.
• Replaces main-thread WebView retries with bounded asynchronous backoff that ignores subresource failures and survives failed-page completion callbacks.
• Initializes the browser accessibility watcher’s datastore connection off the service main thread.
• Adds unit coverage for retry classification, reset behavior, and delay capping.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Android startup] --> B[BackgroundService]
    B --> C{Migrations already queued?}
    C -->|No| D[Queue migrations on IO dispatcher]
    C -->|Yes| E[Skip duplicate migration work]

    A --> F[WebUIFragment]
    F --> G[Load local dashboard]
    G -->|Main-frame failure| H[Schedule delayed retry]
    H --> I[Increase delay up to 5 seconds]
    G -->|Successful finish| J[Cancel retry and reset delay]
    G -->|Subresource failure| K[Keep current page]

    A --> L[WebWatcher]
    L --> M[Initialize RustInterface on worker thread]
    M -->|Ready| N[Record browser sessions]
    M -->|Not ready| O[Drop early events without blocking service]

Loading

_{Reviews (2) · Last reviewed commit: "fix(webui): keep startup retries through..."}

[TimeToBuildBob]

@greptileai review