It's mostly not patches but external mining software. Some of the less good things they do are done because thats whats easy to do outside of the daemon. Not so fun to reimplement consensus rules, so why bother?

The viabtc mining code is public and you can see many such approximations.  E.g. they decide if an incoming block meets the target by the deriving a number of leading zeros off the floating point difficult number rather than an accurate target test.  So if you make a block with a lot of leading zeros but still doesn't meet the target, that code will temporarily switch to it and mine children that will be ignored by every bitcoin node an wallet.   There is no meaningful computational cost in doing the test accurately-- it's just more implementation effort for an external codebase. The fact that it's fairly easy to attack (mining hardware ordinarily returns sub target work, so it would be easy enough to filter that for work that met the relaxed test but not the real target... Presumably they'd fix it once they noticed they lost blocks to it.

One thing to consider is that 2.5% PPS is pretty much guaranteed to go bankrupt per kelly criterion with a reasonable bankroll.  Computationally bounded rationality can be a problem when your system's assumptions depend on rational self interest.

One place where Bitcoin suffers is that it's very stable, and you can keep running old versions. As a result multicryptocurrency pools spend a lot of attention on trashfire altcoins that needs constant emergency fixes, and less on Bitcoin. There have been problems with stuff like newer bitcoin needing a compiler no less than 3 years old but some large miners have much older operating systems.  I don't think there is any real fix except increasing the rate of change-- it's a point I've raised in disagreements that argue that Bitcoin's conservative approach is safer. To a point that's true, but there is a point where being too slow to introduce changes introduces its own risks.

Hopefully taproot marks the end of a one time dry spell and with renewed confidence that clear improvements can be activated without disruption more will be written.