I have my doubts that it is "the finger playing of a piano that
causes these capstans to retreat."  In any case, I've been thinking
that the matter is easily put to the test by placing a small mark on
each capstan with a felt-tip pen.  This would provide a way to see
if they moved under hand playing.

I am not so hide bound (or hide glue bound!) in tradition that I'm
against trying PVC-E (a.k.a. PPCo 320) glue on loose capstans, but even
here in upstate New York, with its fairly significant seasonal changes,
I don't see many pianos with capstans that are that loose.  Of these,
most have had the capstans turned up very high to compensate for
twice-filed hammers (shimmed up hammer rail) and worn action leathers.
In which case, the answer is really to do the necessary action repairs.

Richard Friedman