I cannot physically measure the benefits of this collaborative puzzle-solving. Who cares if it's 100, 500, or 1500 Mk/s? It's the same slow hashing shit, even with all possible fixes. Without something groundbreaking, it's a dead end.
They will work on it for at least two years, and eventually, someone will collect all the prizes using a bot script. It is likely that only resellers of cloud and GPU hosting will benefit.
And if some genius discovers something groundbreaking, do you think they will publicly open-source it? I'm not sure about their safety after that..Which comes to the next point... being slow is what makes them (agencies with three letters) secure. They don't want it to be fast.
There's no hashing for kangaroo, just plain point addition as fast as possible. Which translates to 256-bit arithmetic mambo-jambo.
JLP's kangaroo for example is definitely not fully optimized when it comes to this. But SHA and RIPEMD hashing were heavily studied and lots of hand-written optimized code exists, so the optimization upper bound is likely reached since long ago and nothing better (faster) is likely, just linear speed-up with better hardware. And even if it is, why the hell would one search for a key that, once public, is broken in a matter of seconds anyway? This is a logic fracture...
This is not the case with secp256k1 P + Q, since the longer the private key, the time to break it increases exponentially, but the computations are the same for any length. So it makes sense for scammers to try to convince naive people to contribute the results of some 256-bit mambo-jambo gymnastics (e.g. the bill they pay for running a GPU at full load) for their own benefit (e.g. build a msssively contributed kangaroo jump points database and simply scan for wild-tame collisions). A true non-scam collaborative effort would require that the finder of the key cannot know what the key is, and the owner of the pool cannot know what the key is, but that anyone can verify that the key is correct. So at spend time, a correct prize is distributed. So something like a ZKP protocol for miners.
I think the puzzles are the least problems of agencies, they definitely have access to special-purpose hardware that is beyond our imagination. If FPGAs to break RSA were already built in the 1970s, imagine how things look for them in 2024. We have chips with hundreds of billions of logic gates in our smartphones and call that high-end. How much processing power would, let's say, some processing "chip" that has the size of a basement?