When a PoET replies with its proof late at the start of cycle gap, psm can get into a state where it assumes PoST service is new or registering (that need no proving time). because it returns a READY state. This results in many post-services starting within a short amount of time, creating a "thread fight club" when PoET does reply and proving starts, resulting in poor performance.
Currently psm starts up services based up CG opening layer and current layer, but might need to revisit other CG detection methods, including the classic spacemesh.v1.PostInfoService.PostStates | grep "PROVING"
When a PoET replies with its proof late at the start of cycle gap,
psm
can get into a state where it assumes PoST service is new or registering (that need no proving time). because it returns a READY state. This results in many post-services starting within a short amount of time, creating a "thread fight club" when PoET does reply and proving starts, resulting in poor performance.