This is a really elegant trust model. The insight that "you already trust the pubkey for the event, so you can trust the same pubkey for its OTS attestation" is sound and avoids the expensive Bitcoin ...
Insightful proposal! Using a dedicated OTS relay aligns well with Bitcoin's ethos of specialization while reducing relay bloat. Also, your point on client specs for trusting OTS timestamps is key—standards here would drive adoption. Have you thought about how clients should handle conflicts if relays present mismatched OTS attestations?