For BDP-1 and BDP-2 feeding BDA-1 and BDA-2 in any order, AES has typically been preferred by users. James has also preferred AES with those combinations. With the BDP-3 and BDA-3 combination, James seems to now prefer USB over AES. I think he found it better focused and locked in when I asked him. I am not sure if USB will still be preferred on either the BDP-3 or BDA-3 if used without each other.
Here are some reviews that offer comparisons:
http://www.hifi-advice.com/blog/review/digital-reviews/network-player-reviews/bryston-bdp-3-digital-media-player/https://www.alpha-audio.nl/2018/03/review-bryston-bdp-3-digital-player/7/ (Needs language conversion - use Chrome or google translate)
On my BDP-1 feeding my current DAC, I have preferred AES over (third place) and SPDIF (second place). I've even used the USB with and without Jitterbug.
Between SPDIF and AES, there is a slight difference in presentation that doesn't quite go away. I even controlled for length and wire material (brand) in this comparison. For AES I have 2 feet of Grimm TPR, 10 and 18 feet of Mogami 3173. For SPDIF, I have 2 feet, 6 feet, and 10 feet of Mogami 2964. So I have exact lengths at 2 feet and 10 feet for SPDIF and AES. Cable length can also make a difference in sound.
In direct comparison, I find the SPDIF more upfront, perhaps faster, and exciting sound with more sparkle in the treble. The AES on the other hand sounds more relaxed, slower but in a good way, and meatier in the bass. It's hard to put it into words. In fast switching, you may end up preferring SPDIF as it can come across more detailed. However, I do seem to like AES for long term listening where I forget about tweaking and the equipment.
I am not sure why this difference exists. I asked my DAC company and they weren't sure of why such differences may arise at such short distances. I know James has said that DACs seem to prefer the higher 4V voltage of the AES over the 0.5V of the SPDIF, and that AES naturally offers common-mode noise rejection. Still, I'd be curious to see measurements on how the AES and SPDIF actually measure in comparison to each other and what (if any) differences come out from the analog output of the DAC.