This problem is related to other issues I've seen with using Clusterize on the same dataset in 2 controllers, but in different ways (different sub-bricks, different thresholds, etc.). The somewhat rococo explanation follows:

The issue arises because of the way I implemented clusterization -- the volume to be overlaid is not the chosen OLay sub-brick, but instead is internally diverted to an edited "hidden" volume -- the clusterized thresholded volume. But that hidden volume is attached to the dataset, not to the AFNI controller. So if you are clusterizing the same dataset twice, the hidden volume gets created in A (say), then displayed, then it is destroyed and re-created by B, and then displayed. So far, OK. But the flash procedure, which just alternates between the hidden volume and the "true" OLay sub-brick, may now be calling on the wrong hidden volume!

Fixing this is a problem, since I chose to attach the hidden volume to the dataset structure, not to the controller structure -- and that was because the function that returns a slice for viewing doesn't know anything about the controller structure but is just passed the dataset structure and of course the info about which slice to return. So I have to figure out a way to segregate the hidden cluster-edited volume by controller, and make the slice-creating functions know about it. Not impossible, but requires actual thought to avoid making the code too klunky to understand.

After all, my current goal is now that AFNI survive and thrive another 20 years. Can't make the programs too messy.