April 14, 2026
PRs in Purgatory
The exponential curve behind open source backlogs
One tiny feature, 1 year of waiting—fans yell “fork it” as maintainers drown
TLDR: A tiny Jellyfin feature sat for a year despite approvals, spotlighting how one‑maintainer projects stall under heavy queues. The crowd splits: “fork it and move on,” “fix project management,” or “try AI reviews,” while veterans warn to check stuck pull requests before donating free labor.
Open-source drama alert: a Jellyfin contributor says a small, isolated feature—an on‑screen subtitle timing helper—has been stuck for over a year despite approvals. The maintainer admits the backlog “sucks… for everyone,” but the comment section? It’s on fire. The loudest camp screams: stop begging—fork it and ship. One commenter literally suggests ripping out “90% of code” and rebranding the “faster successor.” Spicy.
Others go big-picture: this isn’t just Jellyfin. Python, Vue—everyone’s drowning. Commenters cite one maintainer trying to hold back a tidal wave of pull requests. One voice calls out open-source’s “original sin”: no one takes project management seriously, so features rot even after green lights. Cue the “dictat… er” jab at leadership styles. Oof.
Then come the “solutions.” Some pitch AI code review to catch the dumb mistakes and speed up feedback, pointing to other projects testing it. Pragmatists suggest “make it plug‑in friendly and write easy, strong tests”—so new features don’t block the main app. Cynics bring the coldest take: before you code, check the graveyard of stuck pull requests. If outsiders’ work never lands, don’t waste your time.
Behind the memes is a bleak truth: when one person reviews everything, queues explode. The author’s queue math—explained in plain terms as “the busier they get, the slower everything moves”—hit the crowd like a chart-topping breakup song. And yes, they’re already calling it the Batch Size Death Spiral. Metal name, miserable vibes.
Key Points
- •A contributor’s three Jellyfin web PRs for a subtitle timeline feature remained unmerged for over a year, despite two approvals and prompt revisions; a 49-line PR saw no human review.
- •Jellyfin web has ~200 open PRs and merges ~20–35 real code PRs per month; about 77 are feature PRs, yet features constitute ~21% of merges.
- •The maintainer acknowledged the backlog and limited review capacity, highlighting reliance on a single reviewer.
- •Using queuing theory (M/M/1), the article explains that near-100% reviewer utilization causes exponential wait times; Little’s Law estimates ~6.7 months average cycle time at current load.
- •The backlog problem is widespread: CPython has 2,200+ open PRs; Vue.js faced unmanageable issue volume; surveys and reports (Tidelift, Ford Foundation) show maintainer scarcity and burnout.