I've got the splitter setup running using two ports on a single SD-22. The delay between the traces is very stable, but relative to on screen cursors it moves around by 1-2 ps. Leo Bodnar commented he thought the low frequency shifting was due to avalanche operation in the heads.
I'm using split cursors on the first peak of the step, but I get similar results if I pick a point midway up the edge. However, I'm using the 5 ps/div range.and 100 fs cursor steps. Using a point midway on the edge seems to let me measure the delay to 200-300 fs with smoothing and 32 point averaging turned on.
This is really just gratuitous playing with a new toy that fairly well blows my mind. I bought four SD-26 heads with 2017 cal dates on eBay for $200 each. When they arrive it will get very interesting as I have an 8 port splitter.
But that requires 8 length matched cables. I'd like to get a crimper for RG402 SMA and 2.4 mm connectors but am having a hard time finding one.
The 11801 FW is incredibly buggy, but when it's working properly it's an awesome instrument. Now I just need an SD-24 and an SD-32 at sensible prices.
Once I figure out how to get machine readable data off this beast I'll be able to do some really cool stuff. At least after a week or two of math and physics ;-)