According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅

Oh actually, we did merge hauntsaninja/mypy_primer#190 upstream, which almost doubles the weight of colour and it appears it now occupies a whole shard by itself, so this won't do much anymore.

Here's a sample run: https://github.com/python/mypy/actions/runs/16173327386/job/45652190197?pr=19399

I wonder if we should revert this then?

Yes, I think so - the zeroth shard dedicated to colour still dominates the total execution time, other shards complete much earlier. However, if #19408 gets through, this will be good again - halving colour check time will make all six shards ~ equal.

Ok, I'll try to review #19408 soon -- if it can be merged soon, we don't need to revert this at all then.