[Bug] MCP server status changes to failed in long-running sessions

Bug Description
an issue on claude code v2.1.85:

MCP server status changes to failed in long running sessions abruptly.

Re-authentication doesn't help. Re-starting the session fixes this.

Seeing this for multiple OAuth based MCP servers.

Environment Info

• Platform: darwin
• Terminal: ghostty
• Version: 2.1.85
• Feedback ID: 3845a69e-ca78-4f82-8976-2b8bbab8e040

Errors

[{"error":"Error: NON-FATAL: Lock acquisition failed for /Users/vivekkaushal/.local/share/claude/versions/2.1.85 (expected in multi-process scenarios)\n    at vp_ (/$bunfs/root/src/entrypoints/cli.js:2794:2168)\n    at yn6 (/$bunfs/root/src/entrypoints/cli.js:2794:1326)\n    at processTicksAndRejections (native:7:39)","timestamp":"2026-03-27T06:47:54.894Z"},{"error":"Error: Tool mcp__rube__RUBE_SEARCH_TOOLS not found\n    at mV7 (/$bunfs/root/src/entrypoints/cli.js:2663:10390)\n    at aA (/$bunfs/root/src/entrypoints/cli.js:681:20998)\n    at Rf (/$bunfs/root/src/entrypoints/cli.js:681:39542)\n    at $YH (/$bunfs/root/src/entrypoints/cli.js:681:50236)\n    at IV (/$bunfs/root/src/entrypoints/cli.js:681:86921)\n    at Eo_ (/$bunfs/root/src/entrypoints/cli.js:681:85884)\n    at lEH (/$bunfs/root/src/entrypoints/cli.js:681:85706)\n    at w7_ (/$bunfs/root/src/entrypoints/cli.js:681:82076)\n    at sH (/$bunfs/root/src/entrypoints/cli.js:681:6489)\n    at XH (/$bunfs/root/src/entrypoints/cli.js:681:4971)","timestamp":"2026-03-27T06:47:58.008Z"}]