Your best chance is if someone can optimize the BLE stack to use less RAM. After all, people have written ipv6 stacks that need a fraction of that.
I imagine the main reason for going with the cheaper chip is that micropython was a pretty late entry to the list of supported coding environments, and the hardware was probably already pretty settled before the port started. Also BLE+4K of ram is probably usable in the other official (compiled) languages, and micropython does otherwise work aside from something that's a fairly advanced feature. So it would have been a tough sell increasing the cost of every board when most students and lesson plans will never use BLE and there are other languages available if they do want it/need it.