Usually, it's a sign of the instance being overloaded. Small instances especially can be prone to this, due to very limited resources. It's recommended to perform two concurrent requests at most when working with a small Splash instance, or scale horizontally by adding more instances of the same size. The latter approach can benefit from the load balancer available to a group of same sized instances, ensuring there's always a working instance under the same endpoint when one of them fails. By contrast, different sized instances do not have a load balancer between them.