Got it! I'm going to expand my earlier suggestion:
TPU Cleanspeak 1.0
1./ When a member posts, the post is parsed and words are checked against the blacklist. Note that the "AI"* can make mistakes, but who cares, because of point 3.
2./ If there are no blacklist words, the post goes up. If there ARE blacklist words in the post, go to step 3.
3./ The post does NOT go up immediately, but goes to PREVIEW mode. The blacklisted words are shown underlined in red. Under the preview/edit window there is a checkbox marked... "TPU Cleanspeak 1.0 spots potentially contentious words. Are you sure you want to submit this post Y/N?". This concept is no different from the "This is an old thread, are you sure you want to wake it from the dead Y/N?"
4./ To post the member must check the box and click post again
5./ The post is parsed once again. If there are blacklisted words BUT the checkbox is selected, the post goes up. If the checkbox is not selected, it goes back to preview/edit mode.
6./ If a member has outstanding infractions, they do not get the checkbox option to post with blacklisted words. Their posts must be 100% clean with the message "TPU Cleanspeak 1.0 does not allow you to post with contentious words until your infraction count is below #"
Bingo! MailMan will now not sleep. He will be caught in the dilemma of wanting to collect infractions, but at the cost of not being able to swear. Poor lad will go nuts with indecision. He will be a changed man. He will be forced to be one, or the other. It might also help people from collecting infractions "on a roll". Once they have an infraction, their follow up posts will be forced clean. How unfortunate! LOL.
*NOTE: Parser "AI" methodology. The "AI" can be as simple as a list of blacklisted words, or could be somwhat smart, catching the typically letter substitutes or double lettering. It could also allow words in context based on the previous or next word, e.g. accepted phrases.
e.g.
1./ During the AI scan, reduce all words to their single letter equivalent. No double letters allowed. Parser would therefore convert "ffuck" and "fuuuckkkk" to "fuck" which would then hit the list and be caught. Note that AI scan would also need to ensure the blacklist was also based on single letters, so that "bollocks" would be "bolocks".
2./ Substitute alternative spellings "1" for "L" or "i", "3" for "E", and "4" for "A" etc. with a soundslike feature.
3./ Accepted phrases "grammar nazi" would be accepted
TPU Cleanspeak 1.0 methodology is hereby gifted to the community based on a single proviso: the words "awsome", "awesome", "unleash", and "world's first" must be on the blacklist. Hark! I have done my bit to save the world from marketing drivel.