Not all endpoints carry the same risk
A read-only status lookup and a face-verify call that can link or unlock an identity don't carry the same abuse risk — treating them with the same rate limit under-protects the more sensitive one.
What tighter limits are actually defending against
Per-identifier rate limiting on face-verify calls specifically guards against brute-force attempts — repeatedly trying slightly different captures or identifiers against the same target until something happens to pass.
Getting the granularity right
Limiting by IP address alone misses attacks distributed across many sources; limiting by the identifier being targeted (a specific NIN or BVN) catches abuse that IP-based limits would miss entirely.