Lesson 8: You can only do log2(n) decisions with Random Forests?

Jeremy states the following with regards to Random Forest early in on in the lesson for Lesson 8 (1:30) :

You can only do around log2(n) decisions, so if there’s a time series it needs to fit to that takes, say, four steps, to get to the right time area, then suddenly there’s not many decisions for it left to make"

Can someone please explain what this exactly means?

Why can I only do log2(n) decisions?
Why are there not many decisions to make after four steps? Wouldn’t it just continue making decisions on other parameters it find are important?