All right, here's the answer. It's a wee bit complicated... 
The TLDR form is: fixing it in all streams in the next build.
In WF5 M130, we implemented some code for how we manage backup information for pattern member features that employ a section. We do this only for new features, to avoid changing existing ones on retrieval.
In Creo 2, we got an SPR, and investigating, found that in the specific case of patterns of multitrajectory sweeps, we should not use this new code. So we made a second change, in Creo 2 M120, to exclude the multitrajectory sweep pattern members from it, again only for new features.
Later in Creo 2, we got a third SPR on a related topic -- this new backup information code was slightly incomplete. This was fixed, but without the necessary versioning check, so it applied to old features as well. This was done in Creo 2 M180. The application of this change to most features seems to have been harmless, but in the case of post-WF5-M130, pre-Creo2-M120 multitrajectory sweep patterns, which shouldn't be using this code at all but were to preserve their old behavior, it could cause the feature to fail where previously it regenerated.
Later (Nov 2015), an SPR was reported noticing this fact. The programmer, presumably busy with a dozen tasks, noted that it could be fixed by flipping the versioning switch from M120, and this was satisfactory to the customer, so they provided a mechanism to do so (the hidden config) in Creo 2 M220. Ideally, they would have said 'while I can patch the data, I can tell this represents a failure to preserve upward compatibility for feature regeneration, and should investigate further', but ideally, we'd have more people doing the work.
A year and a bit later, Wayne enters the story, having encountered the same issue, and files an SPR. The developer investigating notes the cleanup mechanism from M220 gets the job done, and resolves the SPR against that. Wayne however rightly notes that this is an upward compatibility issue, and such things should generally not just be patched.
Last, I read Wayne's post, and agree that it sounds fishy, and do some digging, eventually figuring out the chain of events above. I get everyone's OK to fix it up, and make a fix for: Creo 2 M250, Creo 3 M130, Creo 4 M020, Creo 5 F000, implementing post-facto versioning for the previously-unversioned Creo 2 M180 change, so that models retrieved from older builds will behave as they did the build in which they were previously saved. (And various people are reminded of the importance of upward compatibility and being vigilant about versioning.)
Thus ends the story.
