Eee, I don't think this is a great idea, the DT tests are mega long and we have infra there to shard it. I would think this would fail pretty often because we have to fix DT to work with new DOM types every time in ways that wouldn't be fixable here, e.g. there are packages on DT which exist to provide things that the DOM lib will eventually contain, but then need to be updated once the generated code does actually have things.

Also note that the workflow isn't actually testing anything... Zero tests ran and that's why it appears to be fast.

(In general I think it would be good if we discuss it first before opening PR, to save your time)