In addition to what's already mentioned here about this subject, I think
there's another mechanism which I think reduces the repetition ability
of the notes when the keys aren't locked.  On my upright piano, the keys
are balanced so that they will tip by their own weight, when the pneu-
matic fingers raise the hammers.

When concerning rapid repetition, especially on those Marimba rolls, my
theory is that the whippen collides with the key when the whippen is on
its way down again, and the inner part of the key still is on its way up.
As I see it, this presumably could happen on very rapid repetition levels,
and at certain frequencies, when those two components gets into
disharmony and "phase cancel" each other out.

Thomas Henden