Sorry for my sense of humour. I would say that you do know exactly what is going on. The explaination and reasoning that you gave in your original post was probably spot on.

The differences between the routines is so small that the timings are simply not able to accurately arbitrate between them. Which order they come up in is probably more influenced by whether you moved your mouse, or received an incoming ip packet, or adjusted the volume on your pc, or even cosmic rays rather than any detectable differences in the time they took.

If your on 5.8.1, then it would even be possible that the difference was due to a combination of the randomisation that is being applied to the hashing algorithm and the fairly recent 'stable' sorting methods.

If, (as I assume) the data is maintained internally in a hash, and then sorted for presentation, then the stable sort will tend to leave "equal" elements in their original order, but the original order that keys are returned from a hash will vary from one run to another, hence the randomness you are seeing. Or one possible explanation for it.