Read-only PR review bottleneck brief for nextflow-io/nextflow

[xpinc17]

Hi Nextflow maintainers and community,

I’m building ApprovalOps, a tool for surfacing PR review bottlenecks earlier.

I ran a read-only public-source analysis of nextflow-io/nextflow over a recent 90-day window, using only public GitHub data. There was no app installation, no private data access, and nothing was posted back to GitHub.

I wanted to share the brief in case it is useful for repository maintenance planning. It is intentionally observational: the numbers are historical estimates from public data, not a judgment of the project or a claim that anything is “wrong”.

One thing that stood out: this window shows both first-review waiting time and a few post-approval idle cases. The median first-review wait was 15.9h, while 47 of 219 analyzed PRs waited more than 48h for a first review.

A few highlights from the window:

• 47 of 219 analyzed PRs waited more than 48h for a first review.
• Total observed review-wait was 277.8 cumulative PR-days.
• 11 approved PRs remained unmerged for more than 48h.
• The longest observed first-review waits were 17.8, 14.1, and 14 days.

If this is useful, I’d be happy to share the methodology or run a different timeframe. If this kind of post is not appropriate for this Discussions category, I apologize and I’m happy to remove it.

PR Review Bottleneck Brief

PR Review Bottleneck Brief: nextflow-io/nextflow

• Window: Created 2026-03-08 to 2026-06-06 (UTC, end-exclusive)

• Source: Public GitHub data (read-only, no app install)

• Signal: Sufficient

• Recommendation: This window shows observable review-wait friction worth a closer look. All figures are observed, not guaranteed.

  • 47 PRs waited more than 48h for a first review.
  • 11 PRs sat idle more than 48h after their last approval.

Executive summary

• 47 of 219 analyzed PRs waited more than 48h for a first review.
• 11 approved PRs remained unmerged for more than 48h, suggesting merge-ready work can still fall out of view.
• ApprovalOps is designed to surface aging review items earlier, before they slip out of view.

Observed snapshot

Metric	Value
PRs analyzed	219
Waited >48h for a first review	47
Total review-wait (cumulative across PRs)	277.8 PR-days
Median first-review wait	15.9h
Sat idle >48h after last approval	11

How ApprovalOps helps

In this historical window, ApprovalOps would have surfaced aging review items earlier: an estimated 28% of observable review-wait time, about 77.9 cumulative PR-days. This is a retrospective estimate, not a guarantee.

Review patterns observed

Signals can overlap across PRs; they explain recurring patterns and do not sum to the PR count.

• New contributor in changed area: 94 - The author had limited recent history in the touched code area, which can make reviewer assignment or confidence slower.
• Longer review cycles: 68 - PRs with slower or back-and-forth review activity.
• Quick reviewer alignment: 67 - Reviewers converged quickly with consistent signals.
• Late-stage changes: 46 - Risk-bearing changes arrived late in the review.
• Changed after approval: 46 - New commits landed after the PR had already been approved.

Longest observed first-review waits

Shown as longest observed waits in this window, not typical PRs.

• #7121 Change stdout outputs to stderr, to not break piped output - 17.8 days review wait
• #7090 Normalize TaskPath -> Path in typed process outputs - 14.1 days review wait
• #7113 Fix: Correct fetching of plugin in nextflow plugin CLI command - 14 days review wait

Method & caveats

• Historical estimate only. This is not a live production measurement.
• Review wait time KPI excludes open PRs from the baseline cohort and keeps them only for in-flight diagnostics.
• Aggregate first-review wait metrics may use merge or close time as a fallback when a first-review timestamp is unavailable; the longest-wait examples only include PRs with observed first-review timestamps.
• Read-only public GitHub data may not reflect branch protection, reviewer load, or internal policy in a customer environment.
• Some PRs may have been waiting on CODEOWNERS review state that is only partially visible from public data.
• This read-only public GitHub analysis is observational and may differ from results in a fully installed production environment.