I changed TO92 Q208 2N5400 for MPSA92 and no change in crowbarring.
Before wasting any more time on this I thought I'd remove the crowbar
caps, just in case both were equally bad, so nullifying swap-over test,
but only one PA had enough switch-on instability to trigger either of
the crowbars.
With hindsight these Nichicon VP(M) caps ,2.2uF 50V are much the same as
polarised 2uF 50V caps, suspiciously small.
Simple DVM-R test showed apparent charge time of a 100uF cap, so iffy.
ESR test showed reasonable ESR, but RLC test was interesting.
Initially , for both caps the same, read 1uF then over 10 seconds
dropped to .9uF and continued slowly decrreasing exponentially.
So something in the chemistry gone wrong?, presumably cooked or
something ,with only 50V rating.
Replaced with 2.2uF 250V polyester caps on the other side of the board,
cable-tied down. I know overkill but 100V ones not laying around and
plenty of space on the track side of the output pcb.
Tried 30 switch-ons and no crowbarring and repeated on the other output.