Solution
There are two ways to do this: Software-timed and Hardware-timed.
Software-timed can be achieved by sticking a manual delay between calls the initiate.

You can also do some Hardware-timed by changing the source delay of your first SMU to be whatever you want the timing relationship to be and then wiring the source complete event of that SMU into the other SMU:

Be aware that hardware timing does consume a backplane trigger line implicitly. You can also achieve the above by manually exporting the source complete event to a particular PXI trigger line and then configuring the digital edge on the other SMU to listen to that same trigger line.
Also be aware that in all of these cases there will be significant jitter. The hardware timed method will probably give you ~10us of jitter (per our datasheet, some SMUs are better than others, but this is our typical range). The software timed method will probably jitter by 100's of microseconds.