Here is a case with splitting cells. Initially there are two weak mesocyclone detections associated with the cyclonic south flanks of each cell. You can tell this is a strongly sheared environment by downshear stretching of the cells. Velocity indicates elongated mirror image shear zones on each side of each updraft (like the rock in the stream).

These are sneaky because they can be vertically correlated (since the shear with the obstacle flow is deep enough to show up in multiple cuts). I’m impressed that there is a sanity check (neighborhood check?) in which the number of detections is limited. The panels to the right of the figure above show continued 2D shear with perhaps the beginning of a mesoscylone. Even if there was a well defined mesoanticyclone, the algorithm is not configured to catch the negative shear.

-marfafront