Running a '6700 with 1.3.8 software with DAX link to wsjt-x JT65 software. Using Win7 on a Core i7-920 system with GT640 graphics. We observe JT65 spectra sometimes get "fuzzy" with DAX, but the SSDR panafall is normal. See figure:

This is a wsjt-x spectrum of WWV at 15 MHz showing carrier and sidebands. This run begins at the bottom and goes upward in time. It starts normally, then the "bad" mode sets in. I can always get back to normal mode by exiting DAX client and restarting. You see the red band when there's no data -- I am doing the restart operation. Then follows a period of normal operation. But I can make it go bad again by twiddling the SSDR display parameters - changing black level quickly up and down, presumably loading the SSDR or the firmware. This is reproducible, but note that it only occurs "easily" when my CPU is loaded fairly heavily. (I have another job taking about 50% of CPU in the background.)
I do see this same effect with JT65 and without the extra CPU load, but it takes longer to go "bad". In every case, I've been able to fix it by restarting the DAX client.
I think others have reported something similar, but maybe this adds some useful info.
My hypothesis FWIW: the DAX transmission has a timing-dependent synchronization problem such that the low order bits get scrambled. The high order bits may be OK. The effect is less noticeable with weaker signals. There seem to be two states - "good" and "bad". I can usually flip back from bad to good by using the same trick as I used to make it go from good to bad, but DAX restart is easier.

This is a wsjt-x spectrum of WWV at 15 MHz showing carrier and sidebands. This run begins at the bottom and goes upward in time. It starts normally, then the "bad" mode sets in. I can always get back to normal mode by exiting DAX client and restarting. You see the red band when there's no data -- I am doing the restart operation. Then follows a period of normal operation. But I can make it go bad again by twiddling the SSDR display parameters - changing black level quickly up and down, presumably loading the SSDR or the firmware. This is reproducible, but note that it only occurs "easily" when my CPU is loaded fairly heavily. (I have another job taking about 50% of CPU in the background.)
I do see this same effect with JT65 and without the extra CPU load, but it takes longer to go "bad". In every case, I've been able to fix it by restarting the DAX client.
I think others have reported something similar, but maybe this adds some useful info.
My hypothesis FWIW: the DAX transmission has a timing-dependent synchronization problem such that the low order bits get scrambled. The high order bits may be OK. The effect is less noticeable with weaker signals. There seem to be two states - "good" and "bad". I can usually flip back from bad to good by using the same trick as I used to make it go from good to bad, but DAX restart is easier.