Right now, the checker pool just has some 4 checkers it can use; but the editor can ask for doc highlight, then semantic tokens, then inlay hints, all at once, causing duplicate work and higher resource usage.
We should rethink this, probably splitting the pool into "foreground" checkers and then separate ones for diagnostics, hoping to not create so many checkers at once.
Right now, the checker pool just has some 4 checkers it can use; but the editor can ask for doc highlight, then semantic tokens, then inlay hints, all at once, causing duplicate work and higher resource usage.
We should rethink this, probably splitting the pool into "foreground" checkers and then separate ones for diagnostics, hoping to not create so many checkers at once.