I don't see a reproductive advantage to the plant from this behavior(male reversal limiting female reversal in subs generations), so I'm gonna put this observation down to statistical anomaly. Certainly DJ grew out a whole lotta seeds, but I'd need data sets in the hundreds of thousands to take this as anything else.

There's a strong reproductive advantage to parthenogenesis, which is why it's so damned tough to breed out a tendency to sexual ambiguity. Assuming a gene pool of reasonable size, the advantage will always go to the hermie. In an artificially small pool, parthenogenisis might tend to reinforce negative recessives and thereby self-select out thru poor/failed fecundity. Not sure where the tipping points gonna be there.

Personal experiance has not provided anecdotal evidence ( my favorite non-sequitar) that seed from reversed male displayed gender expression signifigantly skewed from baseline