- Shipped
- June 9, 2026 at 11:49 PM UTC
- Author
- kamo
- Commit
- c5dafe8
/network re-fired fetchNetworks on every background session refresh and hid the whole list on any fetch error, so a transient blip or a stale hasPlatformAdminAccess (wrong-endpoint 403) could silently drop a user's orgs. Background refetches now run quietly and preserve already-loaded orgs; only the initial foreground load owns the spinner + error banner. Also coalesce the duplicate /api/user-info + /api/session/refresh requests that every useUserInfo() consumer fired on mount/visibility/ interval (~30 per page load) into a single shared in-flight request, removing a Redis thundering herd. No time cache — only concurrent duplicates merge, so freshness is unchanged.