I'm wondering how much a Hashcash implementation would help. The approach doesn't seem to be widely attempted. Maybe because we assume spammers are using botnets anyway and don't care about computational cost?