A 4 minute sludge is something that I must admit that I'm out of ideas on how to protect against. The problem is that our load balancing software looks for a certain metric then automatically adds additional web servers into our system if the traffic is bad. Unfortunately this takes 60-120 seconds for the process to detect it then get the servers online during which time that breaking news happens users will get that slowdown.
We've done a lot to stop the site falling over entirely and hopefully people will just be able to cope with that small downtime.
I'd like to point out that right now we have the record number of users simultaneously online, or at least since right after the Bayern game in 2013 if memory serves, and the site hasn't gone down and we only had slowdown for a few minutes so we are certainly making progress on it.
Doing a great job mate, I thought it might have been something as simple as turning a knob up. The old site would have melted by now,