When using the "Shut After Time Delay" transient pump type to model an emergency pump shutdown, changes in hydraulic conditions before the "Delay until shutdown" (such as a valve closing before the shutdown) appear to result in a changing graph of speed (as seen in the Extended Node Data tab of the transient results viewer) rather than a constant speed. Shouldn't the pump operate as a fixed speed until the "Delay until shutdown"?
With the "Shut after time delay" transient pump type, a constant applied torque is assumed, until the time specified as the delay until shut down, when the torque instantly drops to zero. Since the torque is constant, the speed can vary if the conditions warrant.
Essentially HAMMER assumes that no significant changes occur in the model before the pump shutdown occurs (that it is in a steady state until the shutdown). Typically a user would enter a very short time for the "delay until shutdown", such as 5 seconds; long enough to check and confirm that the model is still in steady state.
Your case may go against this assumption, so you will either need to accept it, or change the model to simulate a constant speed instead of a constant torque. This can be done by using the transient pump type "Variable speed/torque". With this selected, choose "Speed" as the Control Variable, and configure the pump Operating Rule with a pattern that directly controls the pump speed. With this you can have the speed constant until the desired delay until shutdown, then gradually drop to simulate torque dropping to zero, based on previously observed speed (graph screenshot above). Note that you also have the option of selecting Torque as the control variable, which makes the shut down easier (just instantly drop to a multiplier of zero at the desired delay until shutdown) but can be more challenging during the time before the shutdown since you need to manually increase the pump torque to maintain the constant speed.
Below is a graph showing an example comparison to the above approach whereby the speed is controlled with an operating rule.
If you are simulating a "normal shutdown" rather than an emergency shutdown, you should consider whether this approach accurately models the "normal" shutdown (is the power instantly cut like with an emergency shutdown, or does it gradually shut down?)
Modeling a pump shut down transient event
Residual flows different for different pump shut down methods
Abnormal pump behavior during normal shutdown of the pump