CNHI finder not deterministic

[SoFiA-Admin]

While fixing a mask data type casting bug earlier, I noticed that the outcome of the CNHI finder in SoFiA does not appear to deterministic. When I ran the CNHI finder repeatedly on the SoFiA test data cube with exactly the same settings, I got a variable number of detections in the range of 6 to 9. Is there an intended element of randomness within the CNHI algorithm that would cause this behaviour?

[RussellJurek]

Not intentionally. I suspect that maybe the algorithm is running into range issues. The probabilities for very obvious detections are going to be so small that this could be happening. As could the comparison to the probability threshold. Hmmm. That's probably the most likely spot.

[SoFiA-Admin]

Why would this result in non-deterministic behaviour? I would have thought that when you run the same code on the same data on the same machine with the same settings, you would expect to get exactly the same probability values unless there is some element of randomness in the algorithm itself.