I watched it again and read the algorithm description part, and I think you're right - D should have been switched to false, as the hand was pointing to it when a cache miss happened.
The behavior is inconsistent with what happens to A and B at the very beginning.
Whenever the hand skips a visited node in search of an unvisited one, it must flip it to unvisited.
If D were given this special treatment every time the hand cycles around, it will forever stay in the cache even if not accessed (D-plomatic immunity?)
