Due to the inductance of the coils, the current through them does not change instantly, rather it ramps up when voltage is apllied, and down again when the source voltage is removed. The result is a sluggish slewing up and down of current rather than a clean switching between full-current/zero-current states.
Current regulation in stepper motor coils is therefore usually achieved by pulse width modulation of the voltage applied to the coils. That is, all of the voltage, or none of it, in pulses. It is not done by regulating the voltage applied directly to the coils.
The result is a coil current which slews up and down a little, oscillating slightly above and below some value, a value which on average has the required RMS magnitude.
In this way, the supply voltage is not a primary concern, although it will influence how quickly current can rise in the coils.
If the controller IC were to use a linear voltage regulator to obtain the exact voltage necessary for some desired current, that regulator would have to dissipate power in proportion to the voltage dropped across it. For example, for 4.2A at 3.8V the regulator would have to dispense with \$24V-3.8V = 20V\$, and would dissipate \$20V \times 4.2A = 84W\$ of power! That's just not practical, and so these devices employ the inductances of the coils to achieve regulation, in a manner similar to how DC-DC convertors use inductors to do it.
If you were to examine the voltage across a coil using an oscilloscope, and if the controller were regulating the current through that coil to be, on average, 4.2A, then you would observe the coil voltage to have an average of 3.8V. This is in spite of the fact that at any given instant the controller is either connecting the coil directly across power supply rails, or to a complete short circuit, or a completely open circuit. Also, if the controller is merely sensing coil current, it doesn't even need to know what exact voltage would result in that exact current. All it does is increase or decrease PWM duty cycle until the current it senses is the current it desires.
The short answer to your question is that the supply voltage will certainly determine the maximum current you may obtain in a coil, but since the controller is using PWM, there's no such thing as an optimal voltage. There's only optimal coil current (for holding, stepping, high mechanical load conditions, low load conditions etc), which is not a direct function of supply voltage. If you wish to achieve maximum torque, you must have a power supply able to provide the current for that condition, and a driver IC able to channel it to the coils. There may be a minimum power supply voltage required to achieve that current, at 100% PWM duty cycle, but any voltage above this (within the controller and motor specifications, of course) will do.