I feel like I'm missing some obvious link here. If changing the name on a legitimate transaction is enough to let it go through, what stops actual Evil Terrorists (tm) from just telling their financiers that their name has changed every three months to stay ahead of specific filters?

If I had infinite resources, IMO, the way to go with this sort of thing is to have no blocked transactions or obvious failures, and only relaying the suspicious transactions for law-enforcement to review. If there are crimes, actually get involved and prosecute them, don't just create weird financial black-holes for innocent people and behaviour. From the customer side, no more stupid disruptions. If there are actual bad actors at play, they don't get any immediate feedback they've been flagged and are less likely to change their strategies and cover their tracks.

If there are too many transactions getting flagged this way and law-enforcement ends up buried under dots to connect, then that's a signaling mechanism for them too-- they've obviously done a bad job of crafting their filters and need to adjust for signal/noise.