The model Claude Opus 4.5 froze while processing the prompt.

[StanGar30]

Describe the bug

The model Claude Opus 4.5 froze while processing the prompt. And it happened 3 times in a row. 3 premium requests were used, 3x each! -_- That is frustrating/infuriating. Please fix things like this so that premium requests are not deducted during a freeze or a bug! Thank you.

Affected version

0.0.369

Steps to reproduce the behavior

No response

Expected behavior

No response

Additional context

No response

[williammartin]

Hey @StanGar30, that sucks, very frustrating. When you say "froze while processing the prompt", can you say more about what state the UI was in, and in what way did it freeze? Did you have to kill the process?

[StanGar30]

Hey @StanGar30, that sucks, very frustrating. When you say "froze while processing the prompt", can you say more about what state the UI was in, and in what way did it freeze? Did you have to kill the process?

Heywilliammartin, i cleared the session using the /clear command, and then, as usual, I entered a rather simple prompt; I'd done this before, and there were no issues. After that, it looked as if the model was thinking and processing the request, as usual, but nothing happened even after 10–15 minutes.

In one of the 3 requests, there was a message: "Model call failed: {"error":{"message":"HTTP/2 GOAWAY connection terminated","type":"connection_error"}}"

I always had to manually interrupt/end the freeze using the "Esc" key.

[williammartin]

Yikes, thanks. Any chance there's anything more in the session logs, which you could find most likely in ~/.copilot/logs/<session>.log where you can get the session ID from the /session slash command. You might also see some request IDs in there that we could correlate back to the server. From what you're describing there was some kind of network hang.

3 premium requests were used,

Also, just to check, was this reported on your billing page, in the CLI output, or both? Sometimes the UI can be out of sync with actual billing since that information isn't exposed to us. Naive hope that maybe you weren't charged for these requests 🙈

[StanGar30]

Yikes, thanks. Any chance there's anything more in the session logs, which you could find most likely in ~/.copilot/logs/<session>.log where you can get the session ID from the /session slash command. You might also see some request IDs in there that we could correlate back to the server. From what you're describing there was some kind of network hang.

3 premium requests were used,

Also, just to check, was this reported on your billing page, in the CLI output, or both? Sometimes the UI can be out of sync with actual billing since that information isn't exposed to us. Naive hope that maybe you weren't charged for these requests 🙈

Unfortunately, I couldn't find anything else in the logs. There's already a lot of my other clutter (or junk) in there.

The /session command also shows "Total usage est: 0 Premium requests", possibly because I've already switched between sessions (or seasons)?

If there was a network failure (or network hang), it wasn't on my end; other CLI terminals were working fine.

I checked right away, and unfortunately, these bugged requests were already counted. To heck with the requests for now (or I don't care about the requests right now); the main thing is that this doesn't happen in the future. Please try your best to fix this. Thank you in advance.

[StanGar30]

Yikes, thanks. Any chance there's anything more in the session logs, which you could find most likely in ~/.copilot/logs/<session>.log where you can get the session ID from the /session slash command. You might also see some request IDs in there that we could correlate back to the server. From what you're describing there was some kind of network hang.

3 premium requests were used,

Also, just to check, was this reported on your billing page, in the CLI output, or both? Sometimes the UI can be out of sync with actual billing since that information isn't exposed to us. Naive hope that maybe you weren't charged for these requests 🙈

It froze again, but this time mid-task execution! -_-

Here's what's in the log: "[ERROR] Each child in a list should have a unique "key" prop. Check your render method of \Box. It was passed a child from aPl. See https://react.dev/link/warning-keys for more information."

[williammartin]

Sad. Have asked around internally to see whether anyone has seen anything similar. No other reports of this yet on this repo. If you grep your log files you should see aRequest-ID, can you provide one of them?

It froze again, but this time mid-task execution! -_-

When you say "mid-task execution" do you mean during a tool call?

[StanGar30]

Sad. Have asked around internally to see whether anyone has seen anything similar. No other reports of this yet on this repo. If you grep your log files you should see aRequest-ID, can you provide one of them?

It froze again, but this time mid-task execution! -_-

When you say "mid-task execution" do you mean during a tool call?

"When you say "mid-task execution" do you mean during a tool call?" - Yes, it looks like it. I confirmed the last action, and that was it. It said it was "fixing" (or "committing") something, and then nothing happened.

Here is the Request-ID from the last time it froze: 00000-e9dea345-68d5-4fe8-8641-98cafeda36e5.

[StanGar30]

The same problem and error occurred again: 2025-12-12T15:35:14.771Z [ERROR] Each child in a list should have a unique "key" prop. Check the render method of \Box. It was passed a child from aPl. See https://react.dev/link/warning-keys for more information. Request-ID: 00000-f1d23902-1e88-4345-82a4-dd488b774d53.

Can you at least give me advice on what I should do or try? Or should I just keep trying until I use up all my premium requests on attempts, and that's it?! I really don't like this.

[ranswife]

I use claude-sonnet-4.5 and his problem happened about 10 times when thinking token reached 15.9KiB (It usually happened when thinking token reached 15.9KiB), sometime maybe like this:

Execution failed: fetch failed

[laeubi]

@williammartin I can confirm that I also have problems with random "hangs" especially on a flaky internet connection. The result is then more or less that copilot is "thinking" forever making no progress but also no error, once has to kill the CLI then because the request for cancel also hang forever (see also #636 for where copilot is not accepting commands in a regular session)!

I can also confirm I see @ranswife response quite sometimes (I haven't seen #595 for a while now).

HTTP/2 GOAWAY connection terminated

This usually happens when there is a proxy server and it no longer wants to wait (e.g. due to longer idle times) this is quite annoying and a problem in other HTTP/2 implementations (e.g. java), so I suspect that in some cases network failures are not recovered gracefully.

[williammartin]

Were you all seeing this on the most recent version, or on older versions too? Any idea when it started happening?

[ranswife]

Only the most recent version. But when I reinstall the older version, it still happened. I think maybe @laeubi is right. I use CloudFlare Zero Trust right now.

[laeubi]

I'm using Version 0.0.358 · Commit f5a8b1e76 but I'm not completely sure how to find out what is the latest and how to upgrade... The readme https://github.com/github/copilot-cli only tells how to install but not how to update neither does it mention what is the latest version...