My vote is to pick the group of pictures that are most important and develop to get the best results for that group.
Consider, also, the results you are likely to get. Place your bet on the horse that's more likely to win.
The way I see it, a three stop push is a long shot. Unless the pictures on that section of film are really, really important, I'd put my money on the sure thing.
The only time I would do something funky like "split-the-difference" development or cutting the film would be if ALL the images are important. By "important," I mean that they are something like crime scene photos, photos of evidence or of a major, newsworthy event.
Otherwise, I vote for what you said. Develop for the majority of the frames that were exposed at the normal ASA of 25 and let the chips fall where they may on the rest.