The VM elasticity code assumed that the service instance is a number, and goes from 1 to 2, 3, ... And uses this assumption to set the instance value when adding or stopping a service instance. In testing/running, observed that if one of these service instances got manually stopped, or when multiple elasticity code managed the same VM, the assumption became invalid, then adding and stopping a service instance failed from then on.

The above assumption is often invalid, need to use a different implementation.