Agreed, if it takes 8 hours to review a PR, then the process is broken and you need to start talking before anyone starts writing code. I'd put the max window on maybe 30 minutes for a PR, otherwise we're doing something else, not a "last pass before merge into production".
I can’t imagine saying, “ah, only six hours of heads down time to review this. That’s reasonable.”
A combination of peer reviewed architecture documentation and incremental PRs should prevent anything taking nearly 8 hours of review.