Don't re-enroll an already hosted drive when opening the sync page

[joepio]

Requested by Joep · project thread

Before: Opening the sync page on a drive already hosted on Cloud Server showed a "Working offline" toast and then "Connected to Cloud Server. Syncing this workspace…" every time.

After: Neither toast appears for a drive that is already on a managed node. A drive that really needs enrolling still auto-enrolls, with the same success toast, and real disconnects still show the offline toast.

The sync page auto-enrolled any account-enrolled drive until the node's proof of hosting arrived, so it re-ran enrollment, and the socket reconnect in enableCloudSyncForDrive looked like a disconnect to NetworkIndicator.

How: new shouldAutoEnroll helper (driveSyncStatus.ts) waits until /server has answered for the current server and skips servers that are already managed; SyncRoute.tsx tracks which server that info came from. NetworkIndicator now reads the ws-disconnected flag per event instead of once at mount. Unit tests for the helper added (2 of 4 new cases fail without the fix). I could not make the toasts show in a browser even with a mocked managed setup, but the unwanted /api/sync-enrollments/challenge request reproduced before the fix and is gone after. Screenshots of the sync page follow in a comment.

---

Generated by Claude Code

[joepio]

E2E shard 3 of run 38060558450 is red on one test: tests/forms.spec.ts "page transitions animate once switched on" (fails on all three attempts, expects the atomic-form-page-exit-up and atomic-form-stagger-in animations and sees none). Nothing in this PR touches forms or animations; it only changes the sync page's auto-enroll condition and NetworkIndicator, and no other shard failed. The failure-state.spec.ts entries in the artifact are the specs' own deliberate synthetic failures. I'm treating the forms test as not this PR's and re-running the failed jobs once; if it fails again I'll look at it as this PR's problem.

---

Generated by Claude Code

[joepio]

Landed in develop through the batch #2210 (squash f39e460), tested together on Mancave run 5204.

---

Generated by Claude Code