Yes, Ken.  Re-squaring the edges (two adjoining) does help sometimes, but not always.  I suspect as you say it's due to rounding of the edges. The etch would of course preferentially attack those right angles.  The mode of oscillation is sheer-thickness in an AT cut, which means the thickness determines the frequency but the blank moves laterally (via shear force).  If something mechanically interferes with that shear, the Q will go down.  This is why the plates of the crystal holder have slightly raised corners.