I think you are making an assumption regarding availability vs price vs demand. I would think that given the sort of capacity Amazon has, at some point of time the cost of a spot instance would be the same as or higher than the cost of a regular ec2 instance, at which time it would make more sense to just get a regular instance.
Per my understanding the attraction of spot instances is that you can get short-term instances at lower rates. Even as I say this I am making a big assumption that Amazon's spare capacity is much larger than demand for sport instances.
Finally slow ping times could be caused by any number of reasons, attributing them to one particular cause based on a coincidental date might be a bit much.
Even if EC2's RAM is all allocated that wouldn't necessarily lead to network congestion, since network usage is workload-dependent. Conversely it would be possible for EC2 to experience network congestion even when its RAM is not full.
I don't know the virtualization system they're using. Could some of the (partially virtual) network stack be software based, and thusly respond to memory pressure?
Per my understanding the attraction of spot instances is that you can get short-term instances at lower rates. Even as I say this I am making a big assumption that Amazon's spare capacity is much larger than demand for sport instances.
Finally slow ping times could be caused by any number of reasons, attributing them to one particular cause based on a coincidental date might be a bit much.