The problem here is the pressure of the time limit which is only compounded by starting off easy in the beginning where there is more time and then ending with the hardest problems which take the longest to solve. Then the question as was raised in the original point 4 is what are you really trying to interview for - speed or the ability to develop a solution in a collaborative manner given a minimum expectation of intelligence?
You can start with a higher-level discussion, such as this:
"Tell me about a time when your team's site/service went down at a critical time (like 10am Monday morning, or maybe it got "slashdotted" on launch day), how you determined the root cause, and how service was restored."
Lots of people are good at BSing such discussions. It's a lot harder to BS with code. Why waste time with a coarse filter when I can start immediately with the fine filter?
