[BUG] Claude Stalling all the time

Resolved 💬 3 comments Opened May 2, 2026 by cc-nogueira Closed May 6, 2026

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?

## Streaming stalls — 3 occurrences in ~7 minutes of active conversation

Claude Code version: 2.1.126
Model: claude-sonnet-4-6 (Sonnet 4.6)
Platform: Linux (Fedora 43, kernel 6.19.13-200.fc43.x86_64)
Shell: zsh

### Symptom

During an active editing session, response streams hang mid-flight for ~15s–60s before resuming. Killing the request with
Ctrl-C and resending "continue" recovers fast (~3s TTFB), suggesting the stall is on the streaming side, not the
prompt-encoding side. The pattern matches lastChunkAgeMs ≥ 15000 debug warnings.

### Debug log evidence (3 stalls captured in ~/.claude/debug/<session>.txt)

2026-05-02T18:18:28.508Z [WARN] Streaming stall detected: 54.4s gap between events (stall #1)
2026-05-02T18:18:28.518Z [WARN] Streaming completed with 1 stall(s), total stall time: 54.4s

2026-05-02T18:21:05.742Z [WARN] Streaming stall detected: 56.9s gap between events (stall #1)
2026-05-02T18:21:05.750Z [WARN] Streaming completed with 1 stall(s), total stall time: 56.9s

2026-05-02T18:25:28.695Z [WARN] [Stall] stream_idle_partial lastChunkAgeMs=14999 bytesTotal=666 idleDeadlineMs=300000

Notable details:

  • Stall #3 fired after only 666 bytes had streamed → response size is not a predictor.
  • Time-to-first-chunk on every API call was normal (~3s, including the calls that later stalled), so this is not a cold-cache

or connection-establishment issue.

  • No [ERROR] lines in the debug log; tool dispatches all complete in <50ms.

### Affected request IDs (for correlation)

b1423c34-1ebd-4a54-ba29-188e70dac76b (stall #1 stream)
fc357338-baec-4137-b639-19466993e32d (stall #2 stream)
9070c805-c488-4310-b759-8fd874a76903 (stall #3 — caught live as stream_idle_partial)

### What I see in the log that does NOT cause it

  • effortLevel: medium, autoDreamEnabled: true, agentPushNotifEnabled: true — none of these correlate with the stall

lines.

  • The 297 Fast mode unavailable debug-level lines are unrelated chatter (Fast mode requires /extra-usage, which I haven't

enabled).

  • Tool dispatch traces ([Stall] tool_dispatch_*) all finish in <50ms — those [Stall] lines are normal telemetry, not actual

stalls. The real signal is Streaming stall detected / stream_idle_partial.

### Workaround

Ctrl-C + retype "continue" recovers because the prompt cache stays warm. Lossy though — partial output is dropped.

What Should Happen?

Claude should answer.

Error Messages/Logs

Steps to Reproduce

Cannot deterministically reproduce — appears intermittent on Anthropic streaming infrastructure. Frequency in this session: 3
stalls in roughly 7 minutes of active back-and-forth (after /debug was enabled; many earlier stalls in the same session
predate logging).

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.126

Platform

Anthropic API

Operating System

Other Linux

Terminal/Shell

Xterm

Additional Information

_No response_

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗