We already have ‘alist->mapping’ where earlier keys take precedence. The rationale for this choice of precedence is to use the same convention as ‘alist->hash-table’ from SRFI 125.
The rationale for that is compatibility with `assoc`, which searches no further than it has to and therefore is earliest-first as a natural result.
IMO, alists are insufficiently appreciated. I've seen code that used hash tables for maps that would never grow past 20-30, a point at which the overhead of a hash table probably dominates linear search.