For example, I have an autonomous tank project. It's got both a Raspberry Pi Zero W and an Arduino Mega 2650 board in it. There's like two GPIO wires that go from the Pi over to the Arduino to send signals to tell it how to steer. Each computer has its own power supply.
But then I notice that if I turn off the Arduino power supply, the LEDs on the Arduino are still on (just faintly). So the Pi is "sinking" 5V over to the Arduino via the serial connection I've created. In electronics design, if power is flowing from one thing to another we say that it sinks power.
Okay, so that's the cause of your undervoltage.
Long-term? Honestly, I don't know of anybody who provides a "smart" USB serial cable which includes an unexpected reverse-direction diode inline with that 5V. Typically, the Type-A side of the serial cable is the host side (imagine a PC computer). The manufacturers all think: plenty of power, let that side of the communications provide 5V to the other side. But in the case where the host is a small single-board computer and the "other side" is the beefy stepper-driving controller board, then this is just unexpected from the standpoint of the people who originally designed USB. You can't have the Pi sink power to the controller.
Honestly, I should make and sell smart serial cables for this industry. The 5V signal (as seen by the Pi from the controller) is supposed to be necessary so that the Pi then knows to create a device in the device tree for it. All that makes the software stack happy. In reality, though, we've got people here who are doing it [the tape thing], it solves their problems and everything works.
If you're handy with soldering, you could buy a good-quality serial cable (with internal metallic shielding or with an external ferrite core), cut off the Type-A end and solder on a replacement, leaving out the 5V power wire. I assume that you had fully tested this with tape first before deciding to do something like this.
An alternate version would be to solder inline a diode for that 5V pin. You'd orient the diode so that current can't flow from the Pi to the controller board. (But in theory, the controller board could present 5V to the Pi which is technically part of the USB handshake.)