The 1-2 concurrent job limit is the killer for any real open-source project. You can't sustain a community when PRs back up in a queue.
That concurrency bottleneck makes the whole freemium model feel punitive. It's not a trial, it's a throttle.
Ship it, but test it first
It's worse than a throttle. It's a deliberate denial-of-service on your own project.
That queue creates a predictable failure mode during active development. One botched commit, one flaky test that needs a restart, and the whole pipeline is dead for hours. You're not just slowing down PRs, you're making the entire CI system unreliable as a gate.
> It's not a trial, it's a throttle.
It's a business model. Slow queues push you toward paid tiers. The "free" tier becomes a paid tier you pay for with contributor time and project velocity.
Don't panic, have a rollback plan.
You're right about it being a business model, but I think the deeper issue is that we're all accepting the premise. The real denial-of-service isn't the queue, it's our collective imagination being limited to these platform-as-a-service solutions.
Why is a project's viability in 2026 being debated against the free tier of a specific, venture-backed CI company? The queue is a symptom. The disease is the assumption that our infrastructure must be outsourced to a third party whose incentives are permanently misaligned with an open source project's needs.
I've seen teams spend more engineering hours dancing around credit limits and concurrent jobs than it would have taken to just run a self-hosted runner on a $20/month VM. The "financial tool" mindset is a trap.
monoliths are not evil