Key findings
- GILD counted 3,781 public Seer pull requests across 543 repositories. Teams merged 1,681, or 44%, and closed 960, or 25%, without merging. Another 1,140, or 30%, were still open on October 12, 2026.
- 926 of those open pull requests were more than 30 days old.
- Ten repositories account for 34% of the pull requests. Without them, the decided merge rate is 63%.
- Across 107 repositories with at least 5 decided pull requests, the median repository merged 75%, and 33 merged fewer than half.
Teams merge most of the Seer pull requests they act on, and ignore many others. Of the 2,641 public Seer pull requests that teams decided on, they merged 64%. About a quarter of everything Seer opened sat untouched for more than a month. Give Seer's pull requests an owner before you widen automation.
Why count pull requests
At general availability in June 2025, Sentry reported more than 38,000 issues analyzed and 94.5% root cause accuracy, and published no method for either figure. Public pull requests are a record anyone can check. The team that owns each repository decides whether to merge a pull request, close it, or leave it open.
A merge does not prove the fix was right, because a team can edit a pull request before merging it. A close can mean the fix landed another way. Each outcome still records what a team decided about Seer's work.
What happened to them
Teams merged most of the pull requests they decided on. Most of the open ones had waited more than a month, which suggests Seer opens pull requests faster than teams assign owners to them.
| Outcome | Pull requests | Share | Share of pull requests |
|---|---|---|---|
| Merged | 1,681 | 44% | |
| Closed without merging | 960 | 25% | |
| Still open | 1,140 | 30% |
GILD count from GitHub search, October 12, 2026
Results vary by team
Merge rates vary far more between repositories than over time.
Teams merged 66% of the decided pull requests Seer opened in the first half of 2025, 60% in the second half. They merged 67% in the first half of 2026, and 63% since July 2026.
- 107 repositories had 5 or more decided pull requests.
- The median one merged 75% of Seer's pull requests.
- 50 merged 80% or more, 33 merged fewer than half, and 7 merged none.
What to do with this
Assign an owner to Seer's pull requests and a stopping point to each project. Review open pull requests weekly, and close stale ones with a reason, so the merge rate reflects the quality of Seer's fixes.
Your private code, internal conventions, and telemetry can move the rate up or down. During a trial, track how many Seer pull requests your teams merge, close, and leave open on your own projects.
Methodology
GILD counted public pull requests on October 12, 2026 through GitHub's issue search API, from March 2024 to that day. It queried two GitHub apps, seer-by-sentry, Seer's deprecated original app, and sentry, the integration that has opened Seer's pull requests since late 2025. It excluded the getsentry, sentry-demos, and codecov organizations and repositories owned by public getsentry members. GILD split queries by date to stay under GitHub's 1,000-result limit. GitHub search does not reach private repositories, where most Seer use happens. GILD counted every pull request from the sentry app as Seer's without opening each one.
Products in this piece
Sources
- Public source
- 2
- Official documentation
- 4