I absolutely agree that we should explore adding sort descriptors to the standard library! In addition to allowing you to compose multiple search criteria, as highlighted in your Stack Overflow answer, sort descriptors also streamline and eliminate errors in sorting based on a single property. We've had a variety of pitches over the years for this feature, and I look forward to working together to bring something compelling to the Swift Evolution community.
That said, sort stability is an important question that can be considered separately at the current time, without waiting for more exploration. In addition to the reasons described in the proposal (it would be disruptive to change even un-guaranteed behavior at this point, and brings benefits to those who would like to rely on sort stability), a stability guarantee is important to have for future work.
Simulating a composed sort is demonstration of stability, but the semantic guarantee is really about preserving a collection's preexisting structure. In your example of sorting by last name and then first name, whether by multiple searches, a carefully written predicate, or by some future sort descriptor, stability preserves the existing ordering of other fields in the data. That is, if the contacts array is already sorted by creation date, a stable sort preserves that order for contacts with exact matches across both names.
Thanks for your feedback!