This seems to be a common problem so I thought I would report on my repair.
Apparently fixed by replacing the U500 hybrid with a known good one.
This is my main scope. In 2015 it received new caps, repaired SMD A5, and new cal. memory. It worked flawless up until a few months ago. At that point it started failing the power on self test with “Test 05 Fail 40” messages.
I set the scope aside dreading trying to find what I assumed would be another damaged trace on A5. After a while, I pulled it off the shelf and started testing it. Of course, it passed all self-tests.
So I put it back in service and waited for it to fail again. Eventually, it did.
This time I carefully measured the power supplies and checked that the DAC was basically functioning. Everything was okay. Before trying to troubleshoot the A5 sample-and-hold circuits I thought I would try an easy experiment. I swapped the U500 hybrid between a 2430A and this scope. Surprisingly, both scopes now work and pass self tests.
I don’t have a good explanation as to why this worked (or indeed if it has — only time will tell). Perhaps the hybrid is marginal and the 2430A uses it under slightly different conditions. Perhaps an issue with contacts — although I had already tried cleaning the contacts and reseating it.