I would certainly agree with Gerry - run the drivers at 68-70V, and go to parallel connection. Series connection is fine but is more suited to lower-speed operation. One of the important things about steppers and speed is to look for low inductance. In effect, this is a kind of electrical inertia that slows the rate at which current can build up through the motor windings and as it's current that gives driving torque, current and fast change of current is what you want. Series connection adds the coil inductances together; parallel connection halves the inductance. Much better!

I'm a bit surprised as well that you are using NEMA 34. Given that the original motors were 2Nm or so (were they steppers or servos?) I would have guessed that you could get away with NEMA23 4Nm. Smaller, lower inertia, lower inductance. If the NEMA23 has enough grunt, it will typically give better performance than a NEMA34, which is contrary to expectation but it can be true. Max useful speed is about 900-1000RPM, which would give close to your nominal 4000mm/min with those ballscrews.

I would have slight reservations about a SMPS as well. These can current-limit under peak load conditions, just when you need the current most. A simple linear power supply with big capacitors on the output can cope much better with peak loads in this situation. People use SMPS quite happily, but they need to be well-specced to make sure.

I'm a little confused about your microstep settings. Again, Gerry has given the usual advice - 8 to 16 microsteps is generally a good compromise between resolution, smoothness and torque. So, x8 gives 1600 microsteps/rev. That's 4mm of movement, so Mach3 should be set to 400 steps per mm.