lovely caps and you could have been having leakage from one of them causing the rectifier to run hotter.
But the failing 2N5886's are after the cap and their failure is probably not connected to a leaky cap.
The supplies need to be calibrated under load in the working conditions that they will be experiencing in the real world. One thing you can do to make this easier is add test points to the changeover unit. (to me this should have been done at the factory)
A lead from each of the inputs to the changeover diodes gives a convenient monitoring point under load. To set them up measure each supply and get them close, final trim can be measuring from one test point to the other. What you want is both of them at the same voltage, exact level is unimportant, and measuring the difference gives the best resolution. You can monitor this over time and see if the regulation is holding or drifting, which is what I suspect. If the difference voltage goes over about 0.7 volts then the current sharing starts to shift from both to one. Hence your problem.
Good luck.
Alan
