Unfortunately, srfi-101:equal? is not necessarily compatible with r7rs:equal?. The latter is free to recurse through records, as Chibi's implementation does, whereas the former insists that two records return #f unless they are eqv?. A PFI should be added to warn implementers of this, and to caution users not to expect whatever the native semantics if they use srfi-101:equal?. So the Right Thing is to indeed to reimplement srfi-101:equal? from scratch.
I just hope there are no other oddball equal? predicates hiding elsewhere in the SRFIs.