WinBox and "infinite" (*1) updates is why I and surely many others like RouterOS, so be careful with that.
It would be interesting to know how much work each part is. It seems to me that once a device model gets going, it shouldn't require much further care for itself. Not counting dealing with past mistakes like 16MB flash with 32MB RAM (hAP mini/lite), where they already had to remove some stuff to keep it upgradeable. But they still sell them, so clearly it makes sense to them. Most stuff should not be device or architecture dependent (when it's either in userspace or using common code in kernel). And when something is, I'd expect it to be mainly the newer specialized things like HW acceleration of different things. In case it's something not already existing in Linux, it could complicate kernel upgrades, but it doesn't happen very often.
Of course there is some extra work, at least with testing different devices, but I'm pretty sure it's done by different people, not developers. I seriously doubt that dropping support for e.g. PPC architecture would significantly speed up development.
As for dropping some features, I don't think it would help much either. Sure, PPTP is deprecated, but it's not like developers spend too much time (if any) on it lately. They spend some on OpenVPN from time to time, but probably not that much either. And it mostly works (if you can live with existing limitations), so why get rid of it? Nobody knows how many customers are happy to have it. Probably more than just a few.
--
(*1) My mipsle museum is very sad for being left out. But realistically, RB1xx were already getting too slow when support ended, so it wasn't too big loss. But RB5xx could still go for a while. It would be also interesting to know what exactly led to this decision, it seemed sudden and unexpected.