Or... maybe in the case of extensions, it's better to leave them in place, but remove all content. So they just have the url field. That way we can always inspect their presence via SQL after the fact.

Unknown modifierExtensions do currently get printed to the console, one per line. So it's a little noisy. But workable for now.

So maybe an MVP of this ticket is: "stop erasing extensions entirely, just erase their content, leave the url in place"

Jamie was nervous enough about including even a whiff of arbitrary extensions (and the more complicated de-id story of "well, we mostly strip extensions", that current thinking is to go back to just logging these on the console.