I've read articles in the past discussing the importance of having timing attack resistant responses for information where knowing the time for evaluation would leak information about the data being processed. More than once the articles mention something along the lines of:
If every response is handled in uniform time it is possible for attackers to know how heavy of a load your server is processing based on response times.
Basically, because an attacker knows your request takes N milliseconds to respond on a normal basis, the attacker can deduce when your server is under a heavy or light load. But I have yet to read a document that shows why this information would be useful.
My question was, does this side-channel leak any valuable information and what would that information be useful for? Also if this information is useful, how to mitigate in a way besides throwing more hardware at the problem?