One effect of a service degradation like you are describing is that people will hit Reload more often when things appear broken, further increasing the load. This can quickly lead to a cascade failure at a tipping point.