A computer-implemented method selectively adjusts a resources addresses cache of addresses of resources used by virtual processors. A first dispatch from a hypervisor dispatches a first virtual processor, and then tracks processes executed by the first virtual processor. The hypervisor caches cache addresses of resources used by the processes after the first dispatch in a resources addresses cache. The hypervisor undispatches the first virtual processor, and then redispatches the first virtual processor as a second virtual processor by issuing a second dispatch. Processes executed by the second virtual processor are compared to processes executed during by the first virtual processor, thus leading to an identification of a level of process utilization consistency. The hypervisor then adjusts the resources addresses cache by selectively clearing resource addresses based on the level of process utilization consistency.