I think you can query S3 by filename, so just save each event as a separate file with name = pubkey:created_at and done?

Or save each event with an id and keep local disk indexes.