At first I thought my issue was with the ground plane, but now it appears the problem is with an initial pose.

I'm using a single Kinect V2 and after calibration (and with the camera at around waist height) I record my t pose and action video. Importing it into Studio and when I get to the t-pose and do a "refit pose" it doesn't put itself into the t-pose but instead gets the legs and body right but bends the arms backwards, even though the arms appear clearly on the video. There doesn't appear to be any way I can adjust this, and tracking from there is a disaster.

This seems really kludgy to me -- the last time I reloaded the video the refit pose instead crumpled the guy up into a ball -- same exact video and same place to fit the pose. Is it this buggy about how it works? One time I *did* get it working right (with another video) and it was great, but if it's not something I can do consistently it won't be any good.

Before hitting the Refit Pose, you first need to roughly align the character with the depth data, using the Move tool.
Please look in the docs for more details:
http://docs.ipisoft.com/User_Guide_for_Single_Depth_Sensor_Configuration#Processing_Video_from_Depth_Sensor

Yeah, it was me being stupid -- for some reason that particular step just wasn't clear in the docs so I just "assumed" (never a good thing) that all I had to do was to refit the pose.

It works just fine when I position it. Sorry for the confusion.