Of course, there is another option, but methinks it would be quite a bit of work (perhaps not practical).

Parse on input, storing the original in a separate table, and the HTML in a "cache" table. The server load would increase a little, as there would be an extra query when a user views their own node, but not horrible.

Of course, that might take a serious restructure of the code, so I'm not sure how good an idea it would be. Just something to ponder...