"Copilot encountered an error and was unable to review this pull request."

[sbelectronics]

Select Topic Area

Question

Copilot Feature Area

Copilot in GitHub

Body

I hope this is the right place to submit this question. I've been using Github Copilot's code review feature on a paid personal/individual subscription. I'm using it to review pull requests that I myself open against my own private repository. At first it worked fine. But shortly later it started failing 50% of the time with this error. Over time the failure rate increased and it's now failing > 90% of the time. There are runners stuck executing code review tasks for as long as 6 hours before they eventually time out.

I tried submitting a ticket, but the only options for submitting copilot tickets are for "Billing, Signup, or activation" and for "Privacy and data protection.". Github's help agent suggested I file it as a privacy concern, but it's been days and was never answered, so I'm not sure that was the right place.

As far as I know, I have not exceeded any quota, nor have I submitted so many requests as to trigger any rate limiting.

It seems to happen regardless of whether the PR is large or small.

Is there anything I should check? Is this a known issue? The copilot code reviews were a really nice feature before they stopped working.

[Radi410]

Since you're on a Individual sub and seeing those 6-hour hangs, you aren't hitting a "usage quota"you're hitting a Context Timeout. Basically, the Copilot Review agent is getting "lost" in your repo's dependency graph or a massive lockfile, and it doesn't have a graceful way to give up.

The "Zombie Runner" Problem
When a review hangs for 6 hours, it usually means the runner is trying to index files that should have been ignored (think node_modules, dist, or massive .json data files). Because it's a private repo, the agent tries extra hard to "understand" the local context to give a good review, but it eventually hits a recursion limit and just spins its wheels until the backend kills the process.

The Support Hack
You’re 100% right that the "Privacy" category is a black hole. If you want a human to actually look at your stuck runner logs, file the ticket under "Billing" and lead with: "I am paying for an Individual subscription but the Code Review feature is technically non-functional due to 6-hour timeouts." Billing tickets have a much stricter SLA and usually get a human eyes-on within 24 hours.

Immediate Workarounds:

Convert to Draft: Toggle the PR to a Draft and back to Ready for Review. This often forces the backend to kill any "zombie" processes attached to that specific PR ID and start a fresh runner.

Audit your .gitignore: If you have large files or minified assets tracked in Git, Copilot will try to "read" them during the review. Ensure your workspace is lean.

The "Small Batch" Test: Try a 1-line change in a new PR. If that fails too, your account’s specific worker queue is likely corrupted and needs a manual reset from GitHub Staff.

[sbelectronics]

I've found that if I go into actions and start canceling the damn things, sometimes they'll generate output anyway.

[Rajveer-code]

Yeah, this sounds frustrating — especially since it used to work and then degraded over time.

A few things based on similar issues people have run into with Copilot code review:

1. This looks more like a backend/job issue than quota

Long-running jobs (hours) + eventual timeouts usually point to:

stuck workers / runners
failed job scheduling
or internal queue issues

If it were quota/rate limits, you’d typically see immediate failures, not hanging jobs.

2. Try to isolate whether it’s repo-specific
   Does this happen on all repos or just this one?
   Try opening a small test PR in another repo and trigger a review

If it only happens in one repo, it could be:

repo size / history
something in PR diffs triggering failures
3. Check for “stuck” runs piling up

If you have multiple review jobs stuck:

cancel any pending/running ones
then trigger a fresh review

Sometimes old jobs clog the queue and cause cascading failures.

4. Timing / load issues

The increasing failure rate you described (>90%) suggests:

either a service-side degradation
or your requests are getting stuck behind something repeatedly failing
5. Not something you’re likely doing wrong

Given:

happens on small + large PRs
no quota issues
long execution times

This doesn’t sound like a usage problem — more like a Copilot service reliability issue.

6. About support categories

Yeah, this is a gap right now — there isn’t a clear category for:

“Copilot feature not working”

If you contact support again, label it as:

“Other / Technical issue” (if available)
or “Copilot → General issue / malfunction”

and include:

example PR links
timestamps
how long jobs are stuck
TL;DR
Hanging for hours → likely backend issue, not quota
Try another repo + cancel stuck jobs
Increasing failure rate suggests service-side problem
Probably needs GitHub to investigate

You’re definitely not alone — this kind of “works → degrades → mostly fails” pattern usually isn’t user-side 👍

[sbelectronics]

Okay, I have to hand it to both of you guys -- you were right to look at the contents of the repository. Somehow a 20-megabyte binary disk image was checked in. Once that image was deleted from the repository, it seems to have started to work fine now. As to why it seemed to degrade over time, maybe it was a load/capacity issue coupled with trying to ingest that large file. I'm surprised it didn't realize this was a binary file and fail fast on it, or at least generate an error message. Oh well, at least it's working now.