How to control target load in Grinder distributed test?

So I'm using the instructions from http://ift.tt/2AJxxt0 to make a CloudFormation stack for running a Grinder distributed test.

The instructions include a DesiredLoad (tests/sec) parameter that the console/agents try to reach.

My question is: Is there a way to control the test so that once the load is reached, the test will sort of "pause" and run tests at that level for a short period of time?

My goal is to gradually increase the load at regular intervals to see at which load the test starts slowing down. The way it works right now, the load and response time just sort of ramp up really quickly, so it's hard to tell at which load the slow-down occurs.

Should I try increasing the interval between the Grinder worker processes' ramp up? Or should I try implementing a thread ramp up? Or should I do both? Does controlling the DesiredLoad parameter not have the desired effect?

Thanks for your help.