The identity "relation" is misconceived since the syntax of "=" is misconceived as a relative term. Actually, "=" is syncategorematic; it forms (true) sentences with a nonpredicative syntax from pairs of (coreferring) flanking names, much as "&" forms (true) conjunctive sentences from pairs of (true) flanking sentences. In the conaming structure, nothing is predicated of the subject, other than, implicitly, its being so conamed. An identity sentence has both an objectual reading as a necessity about what is named, and also a metalinguistic reading as a contingency about the names. Either way the claim about the subject referent has no extralinguistic content. The necessity of alteridentity
(non-self-identity) statements is "lexical", due to contingencies of the names' reference, much like the necessity of analytic statements, due to contingencies of the predicates' sense, and unlike the necessity of logical truths (e.g., self-identities) whose truth is secured by syntax alone.
Both alter-identity and analytic sentences are readable as objectual necessities and metalinguistic contingencies. Epistemically, alter-identity statements are not essentially unlike analyticities.
"Greece is Hellas"/"g=h" and "Greeks are Hellenes"/"(x)(Gx<=>Hx)" are equally (un)informative; so too for "Azure is cobalt"/"a=c" and "Everything azure is cobalt"/"(x)(Ax<=>Cx)". The real epistemic contrast is between proper names (terms without predicative sense) and terms with a predicative sense (names and predicates of properties). Proper names refer to concrete objects, property names refer to abstract objects. That contrast is metaphysical and thus epistemic.