MCP OAuth tokens expire frequently, requiring manual /mcp reconnect

Problem

MCP integrations (Sentry, Atlassian, etc.) frequently lose authentication due to OAuth token expiration. Each time this happens, the user must manually run /mcp to reconnect — sometimes multiple times before it takes effect.

This is especially painful when running automated pipelines that scan multiple external sources in parallel. If any token has expired, that scanner fails silently and the user has to reconnect and re-run.

Current behavior

1. MCP OAuth token expires (seems to happen within hours/days)
2. Tool calls return 401 Token expired or "trouble completing this action"
3. User must open /mcp dialog and manually reconnect each integration
4. Sometimes reconnection doesn't take effect on the first try

Expected behavior

• OAuth tokens should auto-refresh transparently using refresh tokens
• If a token expires, the platform should refresh it automatically before/after a failed call
• At minimum, surface a clear "token expired, reconnecting..." message rather than generic errors

Impact

Any workflow that depends on MCP integrations (Jira, Sentry, Atlassian, etc.) is unreliable because tokens can expire mid-session.

Environment

• Claude Code CLI on Windows 11
• MCP integrations: Atlassian (Jira/Confluence), Sentry, Slack