Stats requests failing too often in cascade

[haggen]

zKillboard has an aggressive throttling in place in which I can only make a single request every second. Right now I have a regular interval of 1 second, where I dispatch a request on very tick. The problem is when a request takes too long the next one might come too soon and then it gets throttled.

The pattern I've seen is that the first few dozen requests takes about 50ms. But then this time begins to increase with each request, taking up to 1–1.5s. That's when the cascading failures start. Once a request takes more than the, all the following ones fail as well.

• One option is to put 1 second between requests, instead of dispatching in a regular interval of 1 second. i.e. dispatch a request, wait for it to finish, dispatch the next one. But this would really ramp up the time it takes to fill the stats.
• Another option is to apply a dynamic interval between them, increasing as the requests take longer and decreasing otherwise.