Engineering

The case for boring releases

After eleven cadences shipped on the same Thursday in April, we stopped chasing excitement and started chasing predictability — and our maintainers slept again.

When we cut our first long-term-support image, the launch felt like a holiday. The whole team gathered around one terminal in the Cape Town office, watched the mirror sync climb past ninety percent, and cheered when it hit a hundred. Two weeks later a kernel regression on older laptop hardware turned that holiday into a fortnight of hotfixes.

Predictability is a feature, not a compromise

So we did the unglamorous thing. We picked a date — the third Thursday of April and October — and we never moved it. Features that were not ready simply waited for the next train. No heroics, no midnight merges, no "just one more patch." The maintainers who package this work across thousands of derivatives finally had a calendar they could plan a life around.

A release nobody has to worry about is the most generous thing an open project can ship.