Not much if you use a password manager that checks the domain before auto filling and/or use U2F tokens as your 2nd factor.

The threat model isn’t too different from other things that can happen if a malicious user is on the same network as you. The scale would be different though.

Homoglyph Attacks come to mind. Replace legitimate domains like nytimes.com with ones that look identical but lead to phishing sites. (The hacker would have to build those convincing phishing sites as well, of course...)