> they could use this information to skip to user-supplied timestamps correctly
If that information is available, it can be gathered by adblockers to guesstimate how the video has been offset, and then you're back to square one.
If "clean" videos are still available (which they likely will be via youtube premium), you still have a reference of how the video should look, so you can start tagging suspicious sections as soon as they are categorically avoided by the youtube "timestamp fixer".
Sure, it will become harder as the frequency of mangled videos increase, worst case every single user will get their unique rendered video, but I think the performance hit on youtube would be immense, because AFAIK you cannot simply cut together separate video streams, even if they are aligned in encoding, resolution and everything beforehand, because it will break keyframes and create artifacts unless transcoded.
If the frequency is relatively low (say, one video per nation or US state) then adblock efforts can just adapt, the existing sponsorblock implementation could probably still work with some modifications.