For a static page, S3 hosting of the html file is way more efficient.
To continue that particular line of thinking... For the "CPU boost" point you bring up, do you understand the throttling as literally reducing the CPU capacity to zero when the boost is "exhausted"? Or simply dramatically reduced?
For the sake of argument again, let's say the Micro instance that is not being throttled by the Amazon hypervisor can handle 0.5% the sustained load of a i7 2600K. That is, 1,000 static requests per second from nginx.
Then, just by conjecture, let's say the throttling algorithm smashes the CPU allocation to 10% of its nominal maximum, thereby reducing our throttled micro instance's capacity to 0.05% that of an i7 2600K. We can still process 100 static requests per second from nginx. That is still more than enough to handle the few dozen requests per second that come in at #1 on the HN home page.
Like I said, it's theoretically possible that they throttle the CPU to literally zero but that seems extraordinarily punitive.
Would amazon allow that throughput, for a micro instance, at least briefly? Maybe, but I'd just be speculating. This I would guess is a more likely issue than CPU usage.