1) It is possible to mount a MITM attack against knockknock just by intercepting a knockknock packet, changing the source IP, and retransmitting. This is a consequence of knockknock not encrypting the source IP within any payload (which involve only packet headers in the knockknock implementation). fwknop does not suffer from this problem because it encrypts the IP to be allowed within the SPA payload. 2) You mention time synchronization issues above. Well, knockknock uses AES in CTR mode, and this also requires synchronization between the client and server of a different kind. If synchronization were a problem for fwknop, then you must apply this critique to knockknock too. You _can_ disable time sync in fwknop if you really want to, but it was added to prevent MITM attacks.
Here are the design goals that fwknop is developed around:
http://www.cipherdyne.org/fwknop/docs/fwknop-tutorial.html#d...