[BUG] Web sessions hang waiting for assistant output; switching tabs + back often "shakes" them loose

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Environment

• Claude Code on the web (claude.ai/code)
• Browser: <your browser + version, e.g. Edge 130 / Chrome 131>
• Session type: long-running remote execution sessions (multiple PRs, tool calls per turn)

What's expected

After I send a message, the assistant's response (text + tool calls) streams into the UI in real time. When activity is happening server-side (tool calls, webhook events, etc.), I see it shortly after it fires.

What's actually happening

For ~the last week, web sessions intermittently stop surfacing new activity. The UI appears frozen on the last visible message even though the assistant is in fact making tool calls and producing output server-side. Clicking another session in the sidebar and then clicking back into the original session often unsticks it — the missing content appears immediately on return.

Sometimes I get an explicit error: "Disconnected from the remote session. Send a message or refresh to reconnect." Most of the time there's no error message — the UI just hangs silently.

Frequency

Multiple times per session, every working day for the past ~week.

Reproduction

Not deterministic from my side, but it happens consistently in long sessions with rapid tool/PR activity. The session-switch workaround works every time.

Impact

• I lose context awareness during the hang (can't see what the assistant is doing)
• I've stopped reporting it because the workaround is fast and reporting interrupts my flow — so this may be under-reported in your telemetry

What would help

• Either: the UI auto-reconnects/refreshes when the WebSocket (or whatever transport) stalls
• Or: a clearer "connection lost, click to reconnect" affordance that doesn't require leaving + re-entering the session

Happy to share session IDs / approximate timestamps if useful — just let me know what's safe to include.

What Should Happen?

• Either: the UI auto-reconnects/refreshes when the WebSocket (or whatever transport) stalls
• Or: a clearer "connection lost, click to reconnect" affordance that doesn't require leaving + re-entering the session

Happy to share session IDs / approximate timestamps if useful — just let me know what's safe to include.

Error Messages/Logs

Steps to Reproduce

1. Start a session and see how far you get before the Claude blinking logo freezes and no assistant response is forthcoming.
2. When response frozen try clicking into 1-3 other sessions and come back and the context often will refresh.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

today's version (I always install updates when they appear)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

_No response_