I emailed Pavel about the DSD issue. He replied asking to set delays/buffers in software longer. Iam at the max in hqp and in the okto tool.
He did say there is a new firmware soon that will hopefully address the issue.
To clarify, when I suggested reloading the firmware, I was referring to his issue regarding losing memory settings (such as programmed remote functions) whenever power was lost to the unit (power cord disconnect, power outage, etc.).
From what I gather, the DSD popping issue is a known issue for OKTO that is being worked on.
Probably completely unrelated, but when I was putting together my IanCanada i2s transport, it had the issue when playing to my DAC at the time, but using a mini Windows desktop did not. In both cases (Windows and Pi), I used the latest JRiver release. I never figured out if the issue was the IanCanada stuff, or something that was inherent difference between USB out from the mini desktop, as I just moved on since it was essentially just toying around.
I should note, though, that running USB from EITHER desktop or Pi through a Matrix X-SPDIF 2, then to DAC via i2s did not experience popping; however, it was once again a situation of USB converted to i2s rather than i2s directly as was with the IanCanada boards.