I suppose Presonus is as interested in the particular fault pattern as I am, especially if the fault current bypasses a ground break and is thus coupled to the monitor inputs –– without affecting the actual sound generated.
I think it is more likely that the ground lift on the DI box HobGoblyn used was faulty. For example, a solder bridge in the switch (assuming it has a ground lift slide switch). Reason is that if the tingle-causing leakage current was transmitted via the left or right signal line (instead of the unbalanced common line, "unbalanced ground"), I think it should have blown the output drivers in the Audiobox Go.
The last resort is USB isolation. USB isolation does the same as a DI box ground lift, except between the Audiobox Go (or similar USB ADC/DAC) and the Mac. Because it is done at the digital signaling, it has zero effect on the data transferred and thus the quality of the audio, and there is no galvanic or even capacitively coupled connection to the host Mac.
One I personally might consider, is the
Hifime High-Speed USB Isolator v2, although to me the price is quite high – as much as a new Audiobox Go costs hereabouts. You'd also need two female USB-C to male USB-A converters, or two USB-C male to USB-A male cables, too.
I haven't torn one down myself, or even seen its circuit board image, but based on their specs, they didn't skimp on the isolated DC-DC converter –– and that is what matters the most.
For USB Audio Class 1, and USB Low speed (LS, 1 Mbit/s) and full speed (FS, 12 Mbit/s), there are essentially two common chips to do USB isolation: Analog Devices
ADuM3160 and
ADuM4160. They are very easy to use, as they only need a couple of resistors and a few bypass capacitors, with everything needed included in the datasheet. The chip does not pass power, though, so the cheap $10 class full-speed isolators vary a lot in which isolated DC-DC converter they use. Most can only provide 200mA to the device, and require a 40mA or so load for their USB 5V to be within spec.
For USB Audio Class 2 and 3, and USB LS, FS, and High Speed (HS, 480 Mbit/s), there are essentially five common chips to do the USB isolation: Analog Devices
ADuM3165 and ADuM3166,
ADuM4165 and ADuM4166, and Texas Instruments
ISOUSB211. They are only slightly more complicated, requiring a 24 MHz crystal oscillator, and preferably some voltage regulators –– they have some internal, but using external regulators reduces the thermal load within the chip, and is worth the extra cost. The Hifime one I linked to above is supposed to use ADuM4166.
There is nothing special in USB Audio Classes, they are just standard USB protocols for audio. There are similar classes for USB webcams, which is why most don't need any dedicated drivers at all. The only thing that varies between the audio classes is whether they work at Full Speed (12 Mbit/s) or High Speed (480 Mbit/s); I do believe they are all specifically USB 2.0 (and not SuperSpeed).
When picking an USB isolator, it is a Very Good Idea to check which isolator chip it uses, and how much current it provides downstream. There is very little that can go wrong with the data isolation – the datasheets cover it all, even suggesting specific PCB layouts! ––, so the isolated DC-DC conversion is what one should focus on when getting one.
Isolated DC-DC converters are typically galvanically isolated, with no EMI-reducing capacitor across the isolation at all; it is just not needed.
There are quite a few DIY projects and design guides to implement one oneself, too. For example, Mark Harris'
full-speed and
high-speed isolator projects at resources.altium.com, Stefan Wagner's (wagiminator)
full-speed isolator at oshwlab.com, and so on. My own
full-speed one is what you typically get when you buy a <$10 USB isolator off eBay, except with soldered-in USB-A male and female connectors; mine is intended to be soldered in-line in a scavenged cable for minimal cost but sufficient functionality. Metal enclosure for the isolator helps reduce any electromagnetic interference, but should obviously be connected only to one of the grounds (host/Mac) and at one point only; perhaps even through a parallel resistor-capacitor network for limiting ESD and other spikes.
Apologies for the long post.