afaik it's not possible to configure buffer size and sample rate from within a user application without exclusive mode, which actually does matter for the non-exclusive use case. iirc it was even worse where applications' streams would be resampled transparently and buffered, which is absolutely not what you want.

I don't use windows for audio anymore so I can't comment on this in win11, but it used to be that WASAPI suffered unless you set your PC in "performance" mode in your power settings whereas ASIO was unaffected.

And yes, latency matters! For live performance you're looking for < 2.5ms of one-way latency to get a roundtrip of under 5ms. After that point it starts being perceptible to players. This is not a performance floor so much as a scheduling one, and ime windows audio scheduling under dsound/wasapi was always shakey.