I question the need for new messages to help SPV clients find the right path because 1) they seem exploitable to me and 2) all they need is the header chain to find the right path and headers are already short and fast to transmit. Waiting for x blocks again seems to protect them if you assume a majority of well connected miners.
This is a far weaker security model that what is achievable.
If you can point a way to exploit this technique, I'd appreciate having it pointed out:
https://gist.github.com/justusranvier/451616fa4697b5f25f60I guess I'm just expressing the opinion that the current header based proof if work mechanism is working fine so far and is difficult to exploit. I'll take a look when I have time (but you're a smart guy so I'm sure the solution is strong). Its just that in this instance I'm not sure we have a problem yet, and I'm starting to take a if its not broke why fix it view largely out of fear that an unneeded change might have adverse unknown consequences (but maybe we will someday and so its very useful to have solutions ready)