I usually mergetest a lot of PRs at once, then merge them one by one as I test them.
In current system, this results in a whole lot of tiny, empty releases. This is obviously not good.

I could try discipline and not merge until I test a whole batch, but that sounds tedious and prone to mistakes.
I could also try merging my mergetesting branch onto main and pushing that, but it only works if all of the tested PRs are ready.

Ideally, there would be a way to:

• Delay releases to something like 1 per day, if something was merged this day
• Automatically delete empty releases and append their changelog to the first complete release
• Detect partly-cancelled releases (cancelled when only some of the builds completed) and treat them like empty ones