adaptivity logic questions

[emmacware]

1. edge case question: when during a substep one of SDs gets removed due to multiplicity dropping to zero, the n//2 / n(n-1)//2 scaling factor is recomputed for the remaining substeps with updated n; while it seems OK, what if the removed droplet was actually using the full dt? or vice versa, it was removed during a substep and the earlier succesful droplets (with the full dt) would have a smaller probability?

2. Is probability (and randomness) affected by only shuffling and retesting only pairs that originally needed the substep?

3. In the multi cell diagram, we are confused about why cell 0 goes from 3 pairs to testing two.. did one successfully complete a timestep?

4. General scaling factor confusion

[slayoo]

5. Extend the multi-cell test: https://github.com/open-atmos/PySDM/blob/main/tests/unit_tests/dynamics/collisions/test_sdm_multi_cell.py (so that answers to questions like 3 are addressed in the test)

(analogously, a new test depicting the question from 1 would be worth adding!)

[github-actions]

Stale issue message