Scale down when a high failure rate is detected #3305
Description
- more context in https://apify.slack.com/archives/C0L33UM7Z/p1765245194935909
- problem - in case of e.g., server outage, crawling becomes very fast, which
- contributes to clobbering the target even more,
- risks busting proxies,
- wastes resources in general
- suggested solution - extend AutoscaledPool so that it scales down when
(request handling failures)/(total request processing attempts)grows over a threshold over some time window