Just want to chime in to answer some specific questions tossed my way:
The ordination yes, the centroids no.
It is very low. A small % of variation is explained by the first 2 PCs! Not necessarily a problem, just that season effects (while the effects evident in that plot) are not earth-shatteringly impactful. Makes sense — while you get species turnover by season, many of the same species will be present across proximal seasons.
Yes. Both permdisp and betadisper accept a single "grouping" factor so you just need to test each factor individually if you are comparing to, e.g., a multi-factor adonis test. I think the original paper on the permdisp method explains why multi-factor permdisp is not possible but cannot recall the reason right now...