Empirical typically means derived from experiment or other structured data collection, as contrasted to theoretical or to anecdotal. SRAM themselves might have warranty data that apply - however, that data might have many holes in it, as many consumers might assume that a prematurely worn BB is wear and tear and not subject to warranty, or SRAM might deny warranty and not record it, they might not have good data on mileage and conditions, etc. If you want actual data, asking around on forums might be the best you can do, and this would be very noisy data.
In the absence of data or anecdotes, we can reason from theory. To the best of my non-engineering knowledge, the more widely spaced the bearings and the larger they are, the greater their load capacity, and the longer life they should have. As you observed, the BB86 shell (at 41mm, see this link) is smaller than the 46mm shell of the T47 standard. For some BBs, the bearings can be placed outboard, I.e. just outside of the shell, much like the Hollowtech II cups sit outside of BSA shells. This enables bigger bearings than if you stuffed them into the shell. I have a feeling this isn’t possible in a BB86 shell, because that standard is wide compared to others (86.5mm width, compared to e.g. 68mm for BSA, Pf30, T47). Thus, all else equal, the bearing load capacity for a BB86 DUB BB should be lower than for, say, a T47 DUB BB, because you’d have to use smaller bearings in the BB86.
I don’t know how this translates to practical bearing life, however. Bearing engineers have methods to calculate expected bearing life, which I assume is derived from a mix of theory and empirical data. To use these methods, you need the bearing size, but you also need to calculate the bearing effective load, and that has to combine the radial and axial loads (the latter are side to side loads, e.g. when you’re pedaling out of the saddle, you impose some axial load on the bearings). Obtaining the correct parameters and translating the result to practical terms is probably beyond most non-engineering laypeople.
I will note that whatever theory might say, I see some high-quality BB manufacturers, like Kogel, White Industries, and Wheels Manufacturing, offer BB86 DUB BBs. I would assume these firms had engineers who believed that the expected bearing life was acceptable for this application. Based on this, I would probably not worry too much about using a DUB crankset in a BB86 bike.