/mcp reconnect (all or single-server) always fails with "MCP controls aren't available right now" — persists across window reload, full VS Code restart, and OS reboot

Environment

• Claude Code VS Code extension
• Windows 11
• 11 MCP servers configured (mix of local stdio servers via .mcp.json and claude.ai-hosted connectors)

Steps to reproduce

1. Configure a new MCP server requiring OAuth (in this case, Cloudflare's remote cloudflare-bindings server, added via .mcp.json with "type": "sse").
2. Restart Claude Code so the server is detected.
3. Run /mcp — correctly reports server counts, e.g. 11 MCP server(s): 8 connected, 3 not connected, 0 disabled.
4. Run /mcp reconnect all (or /mcp reconnect <server-name> targeting the specific unconnected server) to trigger the connect/OAuth flow for the pending servers.

Expected behavior
Reconnect should proceed (e.g., open a browser OAuth prompt for servers needing authorization) or otherwise progress the connection state.

Actual behavior
Every /mcp reconnect attempt — all or a named server — immediately returns:

MCP controls aren't available right now — the terminal is still starting up or is showing another view.

Identical message every time, no variation with retries.

Already ruled out

• Other terminal/process contention — killed unrelated stray dev-server processes running in integrated terminals, no change.
• VS Code window reload — no change.
• Fully closing and reopening VS Code — no change.
• Full OS restart — no change.
• /mcp itself (status query, no args) works reliably throughout — only the reconnect action is affected.

Impact
Any MCP server needing an OAuth/authorization step (remote SSE servers, claude.ai connectors needing re-auth) appears permanently stuck "not connected" with no way to trigger authorization via /mcp reconnect.