I am afraid this won't be so simple. With this patch, when you split reversed strip, both sides will be incorrect. In addition to that, old files that used this feature will break so this would need versioning.

The animation and retiming of the VSE is a bit of a rabbit hole, with things like cache not being invalidated on keyframe changes, f-curves not following transforms, and as in this case this mess caused by making code work for both stips with offsets and strips without. Should all of the problems in this area be mapped out in bug reports, or is some kind of refactor/rethinking of this area being planned? In a case like this it seems like it would make sense to split the function up to either serve strips with offsets or serve them without offsets(but still with synced f-curves).

In D12337#321021, @Peter Fog (tintwotin) wrote:

The animation and retiming of the VSE is a bit of a rabbit hole, with things like cache not being invalidated on keyframe changes, f-curves not following transforms, and as in this case this mess caused by making code work for both stips with offsets and strips without. Should all of the problems in this area be mapped out in bug reports, or is some kind of refactor/rethinking of this area being planned? In a case like this it seems like it would make sense to split the function up to either serve strips with offsets or serve them without offsets(but still with synced f-curves).

I don't think you can rethink how these offsets work really. As soon as you change timeflow, you have to follow all the changes that are applied, and reverse them in all retiming tools. It's easier to do without effect strips like speed effect, but there are challenges still.

These issues are known and stem from design (or lack of one) so there is no point in making many reports for same design issue.

As far as plans goes I made design for speed effect, still need to evaluate all possible solutions, then discuss what to do and when.