My feeling is EAP will be considered mostly feature complete within 3-6 months.
And, all SW development efforts after that will focus on things within FSD as it is sold now.
EAP people on HW2/HW2.5 will be easy to deal with in only supporting them with maintenance releases. If I had an EAP car with HW2/HW2.5 I wouldn't expect a lot of improvements aside from better navigation that led to better NoA decisions. That might be enough for those people. My biggest issue with NoA is really about the failure of the system to know when I need to be in a specific lane. So I think people with EAP will be okay with Tesla simply fixing a few issues. Hopefully they'll also fix the lane-change hiccups when it aborts a lane change for no reason.
As to when the code splits between EAP and FSD? It remains to be seen because it entirely depends on when HW3 is released. It's certainly going to require a bunch of different code as it has a different SOC, and can run a much more complicated neural net.
My prediction is most EAP owners will opt for an upgrade to FSD. Where Tesla will maintain the $2K pricing for 6-12 months to allow that to happen.
HW2.0 owners are the hardest to predict what will happen. I can see Tesla doing a partial upgrade (of just the HW3 computer), but they might also simply say "Sorry, no FSD for you" and refund the money if it's too difficult for them to do all the upgrades (sensors, Radar, wiring, etc).
My biggest concern is Tesla will try to do a water-downed FSD that barely meets the requirements to get away from having to update the car.
I'd advice anyone thinking about spending $2K for the FSD upgrade to hold off until it was known what was going to happen HW wise.