I have a slightly different theory on the pull > push thing, though it'd be interesting if a repairer or two could confirm that bias really exists, on a larger sample size?
i was playing mostly for Morris in that era (by coincidence am back to that this very evening)

These were D/G tunes on a D/G box, and played (I hope) suitably "squarely". I think that when I was on a tonic chord I'd commonly bring in adjacent buttons for a bit of harmony, and indeed volume as the Lllly is single reeded. Playing on the (sub)dominants here was less scope, and I was also learning to cross row a bit, mainly a single reed in play? Add that you can get more force into a pull and I think this'd explain differential stress.
Though little of this applies to PA

also I was learning to bend notes a bit, and those ones didn't go? AFAIR it was the F# and A on the outer row. However these did get some abuse in eg
Idbury Hill