I have reviewed this document as part of the security directorate's  ongoing effort to review all IETF documents being processed by the IESG.  These comments were written primarily for the benefit of the security area directors.  Document editors and WG chairs should treat these comments just like any other last call comments.

The summary of the review is Has Nits

Compared to my last review (-09) a lot of text has been removed, reducing the technical details and giving a briefer (but arguably cleaner) overview.

I do still have a personal preference of dropping CurveCP and MinimalT as I don't really think these are anything but abandoned research items. I have never seen or heard of these being deployed. I would be more tempted to list openconnect (draft-mavrogiannopoulos-openconnect) which actually does see quite some deployment. If the intent is really to show different kind of API's, than there are other esoteric examples that could be included, such as Off-The-Record(OTR) or Signal, which are mostly encrypted chat programs, but with the ability to generate session keys for encrypted bulk transport (eg for video/audio or file transfer)

Section 5.1

 "Session Cache Management (CM):" which does not include IKEv2/IPsec, even though it does have session resumption (RFC 5723). I guess this is because the section limits itself to "the application" that can restart the session. But for IKEv2/IPsec the same could be said. An application that after a long idle period sends a packet, triggers a kernel ACQUIRE, which triggers an IKEv2 session resumption. WireGuard does not have this capability as there are no API/hooks for packet trigger tunnel events (AFAIK)

"Pre-Shared Key Import (PSKI):" lists WireGuard, but AFAIK it does not support PSK based authentication - only public key based authentication. While I understand the IKEv2 entry (PSKs are used for authentication of peers), I am not sure the ESP entry should be here. ESP does not "authenticate peers", unless you call "being able to decrypt and authenticate packets" as an instance of "authenticating peer"

Section 5.2

This states "This can call into the application to offload validation.".  This "can" is only not supported by WireGuard, as all of this is happening inside the kernel. Maybe a note could be useful here?

"Source Address Validation" - It is a little unclear why TCP based protocols are not listed here. They (implicitly) do source address validation. Perhaps the introduction in this paragraph can state this more explicitly. Eg "for those protocols that do not use TCP and therefor do not have builtin source address validation ......."