To me it seems all the work of the same spammer(s). In such a case, do some manual intelligence and wrap it up. It won't scale to all forms of spam, but if a simple regex can uncover 250k+ results in 10 minutes, a manual spam fighter can still block millions of pages a day (and warn the webhost, remove these flakey ads from their networks, etc.).

No doubt the recent machine learning hype has given spammers more advanced tools to avoid detection.