It's not clear to me that this is something that should be genericized, nor that providing a generic HPKE to IETF WGs and then tweaking it as needed is any less

It's not clear to me that this is something that should be genericized, nor that providing a generic HPKE to IETF WGs and then tweaking it as needed is any less work than just composing the correct primitives for each WG use case. As it stands already, IETF WGs are relying on properties that, e.g. DH-KEM only provided by accident. In practice, HPKE is not something you can just use off the shelf, and I personally wouldn't try to write an IRTF standard like it is.