When do two Prolog terms marry? (Was: The issue with free speech)

Now I was struggling giving this predicate
a better name. Namely variant_term bootstrapped
as follows:
variant_term(X, Y) :-
    subsumes_term(X, Y),
    subsumes_term(Y, X).
Why does it need a better name? Well because it
is not really the variant, as realized by
for example SWI-Prolog's (=@=):
For example I find:
?- variant_term(f(X,Y,Z,T), f(A,B,C,D)).
true.
?- variant_term(f(X,Y,Z,T), f(A,B,Z,D)).
true.
?- variant_term(f(A,Y,Z,T), f(A,B,Z,D)).
true.
?- variant_term(f(A,Y,Z,T), f(A,B,C,Z)).
fail.
?- variant_term(f(X,A,Z,T), f(A,B,Z,D)).
fail.
The first 3 test cases match what variant/2
usually does. But the last 2 test cases don't
match what variant/2 usually does.
So how can characterize the behaviour of
this weak variant. I came up with this observation:
- A weak variant takes has variables that appear
   in the left hand side and in the right hand side,
   i.e. common variables, only obtaining an identity
   relation.
- Othewise variables that are either specific to
   the right hand side or that are either specific
   to the left hand side, are associated by a
   bijection relation.
I came up with names like "twin", "sibling" for
this relation. But I got also inspired by the
fact that builtins:variant/2 from Scryer Prolog
leaves bindings. So looking for a name similar
like "unify", but only its weak variant and it
leaves a binding trace. My idea is to use "marry"!