MR29v1 - mfplat: Avoid a race condition when releasing a tracked sample.

[Nikolay Sivov (@nsivov)]

Ok, thanks, I think I understand now. It looks like refcount drops in case on uninitialized rtworkq as well, so we probably should do the same, and in fact it does already happen without your patch. Remaining concern is reworking allocator sample lists so stale pointers are never kept. I'm going to look at that.

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/29#note_594