Just tested this with the following routing:
KPA direct out to analog instrument input of my audiointerface recorded to one track.
KPA SPDIF out to SPDIF input of the same audiointerface recorded to separate track.
(As I understood from a post by Mr. Kemper on a separate thread there is no AD/DA conversion inside KPA when using the direct out, it just passes the analog signal through).
The difference between the simultaneously recorded signals on tracks was around 180 samples which to my understanding would be around 4 ms at 44.1 KhZ (the direct out obviously being 4 ms earlier). Adding to that around 1 ms AD conversion for the analog instrument input of audiointerface (that is missing from the 4 ms latency between tracks) would make the latency of KPA measured from guitar input to track about 5 ms. Not that much, but still different from 2 ms. Maybe/probably this test has a flaw in it - I'm not that familiar with these matters.
Anyway I’ll get back to playing guitar as this 2 or 5 ms really does not matter to me feel wise at all. I’m a very “happy Kemper” and more amazed by this unit day by day. But my tech-nerd side is intrigued – eventhough I try to subdue it.