Performing a key exchange with post-quantum KEMs is very different to ECDH, which will likely cause vulnerabilities in protocols during the migration to post-quantum cryptography.
One example is that different KEMs can have different binding properties, leading to re-encapsulation attacks if you're not careful.
Some less formal guidance is starting to come out about this, but I've not seen anything on this site about this topic. It would be good to get a developer friendly summary of the different types of binding properties, when they matter, how you achieve them, and which NIST finalist/Round 4 KEMs do and do not provide such properties by default.
