As stated elsewhere, using stride instead of videoflip adds a lot of code and complexity. It means we are now dealing with two complex interacting components instead of two essentially self-contained ones. This change is, at least, not justified.
The only justification I've seen for this change is that it works around a GStreamer bug. This justification should at least be given in the patch subjects, and ideally the bug should be at the very least *reported* to GStreamer, especially considering that it would easily affect other projects than Wine.
A better, and plausible, justification would be that using video stride offers better performance even without the aforementioned bug. This would make sense, since the decoder (or videoconvert) can potentially write the video "already flipped", instead of going through a separate flipping step. Is this in fact the case?