I just noticed that when selecting all in a waveform~ the end point of the selection is not the total length of the buffer~. I am pretty sure I've noticed that with very short files this difference is larger than with very long files (which seems counterintuitive if this were a "feature"). I've been searching the docs and archives but don't see any previous mention of this behavior.
This is throwing off some calculations in a patch when pitch scaling with gizmo~ so that different length buffers~ can playback at the same speeds but at their original pitches without FFT time stretch - however the error is too small for my human ears to notice so it's not really a huge deal and if it were I could work around it with info~.
That being said, I am curious why this is the case and/or if I should report it as a bug (using the proper guidelines of course). Here is a demonstration: