Time limits per question or per questionnaire would be useful. There are really two reasons: 1) Intelligence (and often job performance) is measured by how fast a problem can be solved. Within limits, it doesn't matter, and often a better solution that takes slightly longer to come up with is better overall. However, for a problem which is bounded in difficulty (such that a normal job applicant willing to put 2 hours into the interview could answer everything), being able to solve the problem in 24h vs. 2h is not really a good thing.

2) It's a lot harder to "cheat" on a realtime exam, especially on video. The purpose of the interview is to evaluate a candidate, and it's unlikely that candidate's much smarter (or paid) friend would be part of the long-term package, but it's entirely possible someone could be getting assistance during the interview. (it's basically like when a law or consulting firm sends a partner to win the business, and then has summer associates doing the actual work once they're engaged...)

Assuming remote-interview-only, this would be a problem. Also, I view in-person interviews as fundamentally somewhat random. If you really like a candidate on paper (or on video), and he shows up and is close but not necessarily 100%, a lot of interviewers are willing to explain away problems as "he was nervous" or "having a bad day". Once an interviewer emotionally bonds with a candidate, it's a lot harder to be objective -- this is why I like to do early-stage objective/automatic screenings to determine basic competence.