EDIT: Made a video on it. Here’s what’s going on when it’s taking clock input. Every time it snaps, is when I start the sequencer from an external source.
So I had another go this morning. With and without clock input, sampling directly from Tape instead.
When sampling from Tape, it behaves exactly the way I expect it to, and in the way you describe. I’m like, okay. Now it works. I assumed I’d got something wrong, since your description matches how I thought it would work, and since it didn’t (for me) I was using it wrong, somehow.
So here’s what I did now -
When sampling from external source and taking clock, the buffer snaps back to its starting position all the time. Settings are identical as above. It just behaves differently, kind of twitches and shifts as well, as if it’s trying to break free 
a bit of a nervous eye twitch going on there.
So exact same settings and workflow, only difference is external source and clock.
Is this expected behaviour? That the buffer zone behaves differently, depending on if it’s getting clock input or not?
If so, it’s not all that useful to me (but I totally get that it might be for others, so I’m not advocating for change here, just want to understand if I’m getting this right).
At any rate, I’ll wait for the next update and if this is still present, I’ll make a proper repro case and not keep flooding this thread.
