I do not deal with mains voltage. I consider 20 volts "high", because I usually only deal with voltages below 6 volts.
So, I have no idea whether using an USB isolator helps OP with safety at all.
I habitually use the cheap ADUM3160/4160 -based USB 1.1 isolators (12 Mbit/s max.) from eBay, those at or under 10€ / USD $10 apiece; they're direct implementations of the datasheet application example, and only differ in what isolated DC-DC converter they use. They are perfect for use with USB-to-UART converters, as well as microcontrollers having native USB ports at USB 1.1 Low Speed or Full speed (12 Mbit/s). You need ADUM3166 or USBISO211 or similar USB High-Speed (480 Mbit/s) for USB 2.0 High Speed stuff, and even more expensive isolators for USB 3. (Consider gigabit ethernet instead of USB 3, if you need both isolation and high bandwidth.)
Typically, the DC-DC converters can provide isolated 200-350 mA (from 500mA upstream).
ADUM3160 can easily withstand a couple of hundred volts difference (AC or DC) between grounds, and is rated for something like 400-600 VAC continuous (CSA Component Acceptance Notice #5A, IEC 60950-1: 600 V rms (basic); and DIN V VDE V 0884-10 (VDE V 0884-10):2006-12 VIORM = 560 V peak).
The DC-DC converters vary, but I believe the Mornsun B0505S-1WR3 is typical, providing up to 180-200mA; it's specified for 1.5kV DC isolation for at least a minute, 3kV DC for one second; and is protected against continuous short circuit on the isolated side. It is a viable choice whenever the output doesn't need to be well regulated, and will mostly be used for regulation down to 3.3V, as is typical in most USB widgets. Most do omit the input-side EMC CLC filter (4.7µF + 6.8µH + 4.7µF), though; while it is not a problem for typical desktops and laptops (because they tend to have robust USB port electronics), it can be an issue for tiny single-board computers. If there is a 250 Ohm or larger resistor between the outputs of the B0505S-1WR3, it is there because it requires a minimum of 20mA load on the output for stable operation.
Using an USB isolator between the computer and the microcontroller, or computer and the programmer connected to the microcontroller, with the programmer grounded to the microcontroller ground potential, eliminates the possibility of a ground loop. It means the case when the ground potential differs between the microcontroller and the computer is no longer a problem, up to a difference of a couple of a hundred volts.
Also, even if components on one side of the isolator release the magic smoke, even due to an inductive spike of some sort, its effects will be limited by the isolator to that side, and the other side –– usually your laptop/computer/SBC –– is safe. (Things like having a stepper motor turn another stepper motor, with the latters windings having a connection to USB Vcc and GND, for example. No, don't ask, that is, uh, just an example I thought of. I promise. Although "butterfingers" would be an apt nickname for me, considering the number of components I have, uh, explosively investigated the properties of.)
Again, I can say nothing about the safety, though; I'm just a hobbyist. The main reason for my use of the isolator is to reduce stress, as I know that any results from my own butterfingery is limited to circuits on that side of the isolator, and has nothing to do with biological safety.