I added this to my problem report:
I think a clue may lie in the fact that the first clip remains unaffected.
Perhaps the length from the beginning of the original clip is being taken as a reference somewhere in the code and since this is identical to the start of the audio (both value=0) only for the very first clip, subsequent clips think an offset is already present and erroneously try to compensate it's position when the length is recalculated.