"Today solver pods are started at the moment a request for a new model run is received by the platform. The pod takes a few milliseconds to seconds to start. This time could be reduced by always having pods on standby ready to execute runs as soon as they are submitted.

This obviously comes with an additional costs, since then pods will have to be running and will be idle at times that no requests have been done as well. However this cost might be worth it for certain types of planning where response time is critical."