Cant switch using the app

[FoxyIsCoding]

Im using the app on arch linux and for some reason even tho i have multiple accounts logged in. It doesnt show in the dropdown.
Only way to switch for me is to do tailscale switch account_name

[ep520]

Possibly related symptom on v0.18.9 (flatpak, Linux Mint 22.3 / Cinnamon): for me the account dropdown does show my accounts and switching actually works the switch succeeds at the daemon level, and tailscale status in a terminal confirms the new tailnet is active. But the Trayscale window keeps showing the previous profile's machine list; the peer list never refreshes after the switch. Fully quitting Trayscale (flatpak kill dev.deedles.Trayscale) and reopening shows the correct machines.

All profiles have the operator user set, so it doesn't look permissions related seems like the UI doesn't reload peers on account change. Feels like a regression of #175.

[ep520]

I hit a related account-switching problem and dug into the root cause — posting here since it's the open issue on switching; happy to move this to a new issue if preferred.

Symptom (Trayscale v0.18.9 flatpak, tailscaled 1.102.2, Linux Mint 22.3/Cinnamon, three profiles: two tailscale.com + one Headscale): switching profiles (in the UI dropdown or via tailscale switch in a terminal) succeeds at the daemon level, but Trayscale's machine list keeps showing the previous profile's peers until the app is fully quit and reopened. Peer online/offline state is similarly frozen.

Root cause: recent tailscaled versions no longer deliver runtime NetMap notifications over the IPN bus on Linux. In tailscale ≥ 1.102, ipn.Notify.NetMap is deprecated and only populated on Windows (goosGetsLegacyNetmapNotify); NotifyInitialNetMap still yields exactly one netmap at watch start. Since internal/tsutil/poller.go rebuilds Peers exclusively from notify.NetMap, the peer list is populated once at startup and never updated again — profile switches only refresh State/Prefs, hence the stale list.

The netmap-deprecation-fix branch is the right direction (InitialStatus + PeersChanged/PeersRemoved), but running it against tailscaled 1.102.2 I found three gaps:

1. The watcher dies with 400 Bad Request. The branch keeps NotifyRateLimit in watcherOpts, but tailscaled 1.102+ validates the mask (ipn.ValidateNotifyWatchOpt) and rejects NotifyRateLimit combined with any of NotifyPeerChanges | NotifyNoNetMap | NotifyInitialStatus (NotifyRateLimitIncompatibleBits) — rate-limiting a stateful delta stream would corrupt it. WatchIPNBus then 400s forever in the retry loop and the UI stays empty. Fix: drop NotifyRateLimit from the mask.

2. Profile switches merge both tailnets' peers. On SwitchProfile, LocalBackend.resetForProfileChangeLocked enters NoState and discards its netmap without sending PeersRemoved for the vanished peers; the new profile's netmap then arrives as a full-set PeersChanged upsert. With deltas applied to the existing map, the UI shows old ∪ new peers. Fix: clear Self/Peers/FileTargets/Engine when the state transitions to ipn.NoState or ipn.NeedsLogin, so the next netmap rebuilds from a clean slate.

3. The local machine shows up as its own peer. tailscaled occasionally delivers updates about the self node through PeersChanged (promoted self patches during reconnects). Fix: in applyPeersChanged, treat a node whose StableID matches Self as a Self update instead of inserting it into Peers, and drop Self's ID from the peer map on SelfChange.

With the branch plus these three changes, switching between all three of my profiles updates the machine list live in both directions (UI dropdown and CLI tailscale switch), with no restart needed and no duplicate self row. I have the three commits as a clean series on top of the branch — happy to open a PR if useful.

[DeedleFake]

Is this still happening with v0.19.0?