The DSX satellite looks pretty light. The paper
Design and Systems Engineering of AFRL’s Demonstration and Science Experiments says it masses less than 600 kg. This should give FH plenty of performance. Arabsat went into a transfer orbit that required LEO+2960 m/s. Reducing the mass from 6465 to 600 kg provides a lot more performance. With the usual assumptions (107t fuel, 5.5 empty mass + residuals, ISP=348), about 2125 m/s more, so a total of LEO+5085 m/s available. This should be enough to put the satellites in final orbit, and then actively dispose of the stage. So I wonder why they are not doing that? Maybe they don't want to put the droneship out so far and stress the second stage. Maybe the second stage cannot light 5 times?
The below calculations use the law of cosines: if you have one vector of length a m/s, and you want a vector of length b m/s at an angle of theta away, you need dv = sqrt(a^2+b^2-2*a*b*cos(theta))
So they start with a 28.5
o LEO, and release the first satellites. Then they need about 600 m/s to do a 4.5
o plane change to 720x720x24
o, and release the second set.
Next, assume they go to a 720x6000x24
o, coast to the top, then boost to a 6000x12000x43
o. This is not the most efficient (better to do the plane change at 12000 km, and also better to do some of the plane change with each burn) but it's a worst case. Also it's the shortest time to final satellite release, which may be preferable for reliability.
720x720 -> 720x6000, both 24
o, takes 956 m/s. (7497.5 m/s in LEO, need 8453.5 m/s)
720x6000x24
o -> 6000x12000x43
o takes 2264 m/s. Going 4845.3 at top of transfer orbit, need 6205.6 at bottom of final orbit, plane change of 19
o.
Now they drop of DSX in its orbit. So far they have used 600 + 956 + 2264 = 3820 m/s after LEO.
Now (in theory) they should still have plenty of delta-V left to dispose of the stage. At the top of a 6000x12000 orbit, speed is 4179 m/s. To reduce this to a 0 x 12000 orbit, need to reduce the speed to 3343 m/s, or remove 836 m/s. They should easily have this left over.
NOTE added later: This neglects that first (orbital) burn and plane change are done with more mass, and the mass of the dispenser. So the margin might not be there...