Hi, I started contributing to Airflow in the first half of this year. It is my very first open-source project, and I have learned a tremendous amount from it.
I fully understand that maintainers invest a substantial amount of time and effort into reviewing every PR. In today's era of AI, I've come to realize that code reviews are often far more critical than simply opening PRs. Once I actually started helping with reviews myself, I noticed that doing a thorough review requires significantly more time and mental effort than just picking up and opening a PR. (Even though I currently have the highest number of open PRs in bucket A ,) I completely support Airflow's proposed PR limit policy to address the current review bottleneck. As a student, I have quite a bit of free time to dedicate to Airflow, which gradually led to a backlog of PRs that have stayed open for quite a while. I truly appreciate how valuable maintainers' time is, and I am always genuinely grateful whenever someone takes the time to review my PRs. I believe this policy will give maintainers some much-needed breathing room. Moreover, I think this restriction will encourage everyone to reflect on what makes a truly meaningful PR, and what impactful contributions can be made beyond opening PRs, rather than just chasing PR counts. In short, even though quite a few of my own PRs might end up being closed, I am more than happy to support this initiative. Colten Sameer Mesiah <[email protected]> 於 2026年9月22日週二 上午3:42寫道： > Hi Jarek, > > Thank you for pulling up the statistics. It really helps us understand the > situation better. But I think these point-in-time numbers can sometimes > mask the natural deviations in activity for most contributors. Since you > took me as an example of someone who would hardly be affected by this new > policy, I think we need to look at the maximum number of PRs I had open > over my history of contributing. My memory is not 100% accurate of course > but I recall having 15-20 open PRs at one point. And right around the time > of the summit, I am fairly sure I had 10 open PRs (some in draft though I > am not sure if this policy applies to draft PRs) which then went down to 6 > as the maintainers returned to start reviewing and merging my PRs again. I > do not fault the maintainers of course as I understand the review queue is > going to deprioritised during that period due to intense summit > preparations. > > I would say I’m opening fewer PRs nowadays (unfortunately other > commitments just get in the way even when you really want to contribute as > a volunteer). So the 6 open PRs statistic might have caught me in a low > point. > > I do not wish to make this thread about myself but I suspect other > contributors’ activity might have similar patterns. Periods where they have > lots of open PRs and then periods where they only have a handful. You can > clearly see this in the statistics of people in bucket A who you seem to > categorize as ‘good’ contributors. > > If you are insistent on this policy, I think a 10 PR limit might be a more > reasonable middle ground. It will prevent bots and bad faith contributors > from spamming 50 PRs but it won’t be excessively restrictive. > > On Mon, 21 Sep 2026 at 20:23, Jarek Potiuk <[email protected]> wrote: > > > Correction: Thursday the 24th at 7:00pm CEST. - which is 72 hrs from the > > beginning of that thread, > > > > And one more comment: I thought for a long time about the metrics > "Threads > > vs. PRs". For me this is **embodiment** of community-aware, empathic > > contributor - on the way to become a commmiter. And I **knew** - because > I > > checked it before - that practically **all** committers have a very high > > number here. > > > > I deliberately never mentioned it in our discussions. As we all know, > > anything you set as a goal stops being a useful metric. There are always > > ways to game the system. > > > > That's why I waited for quite a while to accumulate the data and showed > > completely different numbers - and never discussed the thread vs. PRs > idea > > - to ensure this particular metric remained totally unbiased. > > And - just for the record - this is the last time this metric will be > > unbiased—now that we will publicly say we "care" about it. > > > > And here is Claude's analysis on those top-20. And my thinking is that we > > should care about the needs of groups in this way: A > B > C. > > > > ------------ > > > > Three buckets, all 24 (top-19 + the five-way tie at 6 open PRs) > > > > Bucket rule, stated up front so the borderline calls are visible: > > engagement = others' threads per open PR (≥1.5 counts as engaged), > quality > > = all-time merge rate (≥40% healthy). I picked these thresholds to match > > where the data actually clusters - there's a real gap between 1.7 and 0.8 > > on engagement, and between 40% and 31% on the merge rate. > > > > --- > > > > A. Net reviewers — they give back more than they ask for (9 people, 103 > > open PRs) > > > > ┌─────────────────┬──────┬─────────────────┬─────────┬─────────┐ > > │ Author │ Open │ Threads/open PR │ Reviews │ Merge % │ > > ├─────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ SameerMesiah97 │ 6 │ 60.3 │ 350 │ 90% │ > > ├─────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ shivaam │ 6 │ 13.0 │ 55 │ 67% │ > > ├─────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ Vamsi-klu │ 10 │ 10.8 │ 58 │ 38% │ > > ├─────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ Andrushika │ 9 │ 4.6 │ 33 │ 79% │ > > ├─────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ yuseok89 │ 12 │ 3.8 │ 49 │ 77% │ > > ├─────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ steveahnahn │ 14 │ 3.2 │ 40 │ 59% │ > > ├─────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ ColtenOuO │ 23 │ 2.6 │ 70 │ 59% │ > > ├─────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ stephen-bracken │ 6 │ 2.3 │ 16 │ 56% │ > > ├─────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ fat-catTW │ 17 │ 1.7 │ 33 │ 50% │ > > └─────────────────┴──────┴─────────────────┴─────────┴─────────┘ > > > > Vamsi-klu is the odd one: 108 threads and 58 reviews (bucket-A > engagement) > > on a 38% merge rate. Placed here because the review effort is > unambiguous. > > fat-catTW and stephen-bracken are the borderline entries - drop the > > threshold to 2.0 and fat-catTW moves out. > > > > B. Shippers, not reviewers - volume with normal merge rates (6 people, > 115 > > open PRs) > > > > ┌───────────────┬──────┬─────────────────┬─────────┬─────────┐ > > │ Author │ Open │ Threads/open PR │ Reviews │ Merge % │ > > ├───────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ Eason09053360 │ 35 │ 0.3 │ 13 │ 52% │ > > ├───────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ 1fanwang │ 34 │ 0.8 │ 36 │ 40% │ > > ├───────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ rjgoyln │ 27 │ 0.7 │ 34 │ 53% │ > > ├───────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ Samin061 │ 7 │ 0.4 │ 4 │ 58% │ > > ├───────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ jeff3071 │ 6 │ 0.3 │ 7 │ 46% │ > > ├───────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ YAshhh29 │ 6 │ 0.2 │ 1 │ 45% │ > > └───────────────┴──────┴─────────────────┴─────────┴─────────┘ > > > > The work lands—merge rates here are ordinary, some above > committer-adjacent > > levels—but the top three are carrying 96 open PRs between them while > having > > commented on 56 of other people's threads in total. 1fanwang sits exactly > > on the 40% line. > > > > C. Low engagement and low merge rate (9 people, 82 open PRs) > > > > ┌───────────────────┬──────┬─────────────────┬─────────┬─────────┐ > > │ Author │ Open │ Threads/open PR │ Reviews │ Merge % │ > > ├───────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ bujjibabukatta │ 16 │ 0.2 │ 5 │ 11% │ > > ├───────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ anmolxlight │ 12 │ 0.4 │ 12 │ 31% │ > > ├───────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ dheerenmohta │ 10 │ 0.2 │ 0 │ 28% │ > > ├───────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ namanjain24-sudo │ 8 │ 1.1 │ 1 │ 11% │ > > ├───────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ Kunal8954 │ 8 │ 0.8 │ 6 │ 20% │ > > ├───────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ bramhanandlingala │ 8 │ 0.0 │ 1 │ 22% │ > > ├───────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ zozo123 │ 7 │ 0.1 │ 6 │ 17% │ > > ├───────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ GayathriSrividya │ 7 │ 0.0 │ 9 │ 22% │ > > ├───────────────────┼──────┼─────────────────┼─────────┼─────────┤ > > │ nagasrisai │ 6 │ 0.7 │ 6 │ 29% │ > > └───────────────────┴──────┴─────────────────┴─────────┴─────────┘ > > > > bramhanandlingala and GayathriSrividya have commented on zero threads > they > > didn't author. This bucket is 82 open PRs with an aggregate merge rate > > around 21% - roughly four reviews spent per PR that eventually lands. > > > > > > > > J. > > > > > > On Mon, Sep 21, 2026 at 9:09 PM Henry Chen <[email protected]> wrote: > > > > > Thanks Jarek and Vincent for the detailed data and perspective that > > > context is really helpful. Seeing the actual numbers (especially that > > even > > > the top contributors rarely hit 5 open PRs), a cap of 5 makes sense to > me > > > as a starting point. Happy to go with it and adjust later if we find > it's > > > too tight in practice. > > > > > > Thanks > > > Henry > > > > > > --------------------------------------------------------------------- > > > To unsubscribe, e-mail: [email protected] > > > For additional commands, e-mail: [email protected] > > > > > > > > >
