Sessions frequently lock up and get stuck in_progress until timeout

[Cform]

What’s happening

Happier Dev has been locking up regularly for me. A session will work for a while, then the UI/runtime appears to get stuck. At that point the active operation never resolves, and the session either times out after 120s or I have to abort/kill it.

Environment

• Happier CLI: 0.1.0-preview.1771774953.99369
• npm package version: 0.1.0
• Node: v25.8.2
• macOS: 26.3.1

Expected behavior

Tool calls should either complete normally or fail cleanly without leaving the whole session effectively hung.

Actual behavior

A session works briefly, then some tool call gets stuck in in_progress, the session becomes unresponsive, and eventually it times out or has to be aborted/killed.

Relevant log evidence

1. Repeated in_progress / timeout-bump loop

Source: ~/.happier/logs/2026-03-25-20-45-35-pid-35827.log

[08:12:54.696] [AcpBackend] Received tool_call: toolCallId=call_wT5FYBOMxITAndhMfojDyQFY, status=pending, kind=execute
[08:12:54.697] [AcpBackend] Tool call call_wT5FYBOMxITAndhMfojDyQFY is pending permission; skipping execution timeout setup
[08:12:54.853] [AcpBackend] Permission request: tool=bash, toolCallId=call_wT5FYBOMxITAndhMfojDyQFY, inputKeys=command
...
[08:13:02.003] [AcpBackend] Set timeout for call_wT5FYBOMxITAndhMfojDyQFY: 120s (bumped)
[08:13:02.004] [AcpBackend] Received session update: tool_call_update {
  "toolCallId": "call_wT5FYBOMxITAndhMfojDyQFY",
  "status": "in_progress",
  "kind": "execute",
  "hasContent": true,
  "hasLocations": true
}
[08:13:02.004] [AcpBackend] Set timeout for call_wT5FYBOMxITAndhMfojDyQFY: 120s (bumped)
[08:13:02.005] [AcpBackend] Received session update: tool_call_update {
  "toolCallId": "call_wT5FYBOMxITAndhMfojDyQFY",
  "status": "in_progress",
  "kind": "execute",
  "hasContent": true,
  "hasLocations": true
}
[08:13:02.005] [AcpBackend] Set timeout for call_wT5FYBOMxITAndhMfojDyQFY: 120s (bumped)

2. Permission request effectively hangs, then times out

Source: ~/.happier/logs/2026-03-16-15-55-06-pid-40469.log

[15:57:52.698] [OpenCode] Permission request sent for tool: bash (call_function_txjiyum05x38_1)
[15:57:52.984] Agent state updated {"requests":{"call_function_txjiyum05x38_1":{"tool":"bash","arguments":{"command":"once"},"createdAt":1773694672698}}}
[15:59:52.688] [AcpBackend] Tool call TIMEOUT (armed on in_progress) (from tool_call_update): call_function_txjiyum05x38_1 (execute) after 120s - Duration: 120.03s, emitting terminal error and removing from active set
...
[16:01:46.975] [OpenCode] Permission approved for bash
[16:01:46.978] [AcpBackend] Permission response event (UI only): call_function_txjiyum05x38_1 = true

3. Stuck session followed by manual abort/kill

Source: ~/.happier/logs/2026-03-25-19-50-31-pid-32825.log

[20:14:47.940] [AcpBackend] Tool call TIMEOUT (armed on in_progress) (from tool_call_update): call_iCWCKnQhqM9EWg37729RVPcY (other) after 120s - Duration: 120.06s, emitting terminal error and removing from active set
[20:14:47.946] [AcpBackend] No more active tool calls after timeout, emitting idle status
...
[20:44:41.702] [OpenCode] Abort requested
...
[20:45:22.730] Kill session request received
[20:45:22.730] [OpenCode] Kill session requested
[20:45:22.731] [OpenCode] Abort requested

Additional note

I also saw a lot of repeated socket log lines like:

[SOCKET] [UPDATE] Decrypted new-message is not a UserMessage payload; forwarding generic event

Not sure if that is causal, but it seems correlated with the stuck periods.