I think there are two qualities to the problem
1) what kind of authentication takes place to admit out-of-region data into a system which demands self-referential integrity and can't be made to do cross-system references
2) what time limits do we place on the data to require re-validation, so that it doesn't last forever and go stale.
Designing this demands both sender and receiver agree. The prior art, RPSS and RPS-Auth did not achieve agreement both sides: we didn't all agree to run a single cohesive framework.
RPKI (noting Sanders concerns it scares some people) has the huge benefit: all the RIR are doing it, and all the RIR respect each others root/signing trust chains.
And, as I said before, it has time limits built in: signed objects have a lifetime by definition. Do nothing, and data ages out at some point.
Thats why I like it: its commonly implemented, and it behaves the ways we need, for this function.
-G