Every layer added to your audio pipeline will add latency, to say nothing of excessive layering introducing unnecessary compatibility issues and bugs.

I would really like to see OSSv4 adopted as the de-facto standard. It has a simple API, includes software mixing and virtually everything anyone could want. I don't think Linux needs its own audio API just to be different from everyone else.