I think it would be nice to actually see the connectivity structure in the dynamic connectivity example. I'm happy to make a quick PR to do plot_circle and make it look nice.

I think a deeper question is whether the connections are strong enough not to be likely to be observed by chance, which I think is a question people doing this analysis might be interested as the main result from this analysis. Perhaps, since there is a decent number of epochs, a cross-validation approach would be reasonable. Perhaps with a bootstrap approach of shuffling the data in time to create a null distribution (maybe also preserving autocorrelations by just sliding the channels around relative to each other as in this paper https://pubmed.ncbi.nlm.nih.gov/34717794/). Curious to know what others think, especially @adam2392. Maybe this is too much computation for an example also...