That seems like another problem with your approach that doesn't exist with a timed, controlled sequence. In fact, once you start the process you have no exact idea where the pot is set, because you depend on it over stepping the end to finish the operation.
If the direction suddenly changes with a timed application of tone(), you could double the time you waste, over stepping the end. To fix that, requires superfluous logic that would better be leveraged at controlling the steps individually.
With timed, measured pulses you always know exactly where the wiper is (because you can count steps), so when you reverse direction you know exactly how many steps you need to take.
Not knowing whether the travel will complete at the outset, requires that you allocate travel time for the worst case, 63 steps forward and 64 back. If you reset the tone() timer to avoid this, you are not simplifying the logic at all beyond basic discrete step control.
In the case where you reset the tone() timer every time you change direction, you will not be able to perform another operation for an unnecessarily long time, if the wiper barely moved before that happened.