It'd look like automated spam, and they'd block/flag it on the spot.

What would happen to your fb account? Do they only block the account or the spam?

Also could you use ML to be more discrete to avoid the spam flag?

Not sure about ML, but there're several problems to solve before that. The service would need IP address that aren't from known cloud provier, otherwise it's trivial for them to notice and block. Also it'd have to look like a real user, so a full browser (posting directly to their endpoints will get you flagged). Then, each user can't post too much and the IP would need to be recycled as well.

Too much trouble to even start that.

You can rent a botnet of desktops from residential IP.

For example, the Hola "unblocker" C&C people have this service. https://luminati.io/

A service that continually updates your profile data (well the non-invariant type), adds/removes random "friends" and the like would be interesting.

Do it progressively enough, and when would FB know the difference?