As I kind of suspected, Analog got back to me and informed me that the ADE7763 was a very mature product and that the reversal of IRMS and VRMS was never reported. Now unless NOBODY has ever tried to read these registers, than the problem must be mine.
Thanks to Analog for taking the time to talk to a low volume user.
Time to go back to the lab and start a very careful analysis of the problem.
After a hour of careful testing, I found that the register values are not reversed, but rather they do not return the correct value the first time you read the register. If you read the VRMS for a while and then read the IRMS register, the first read is the VRMS value, the second and subsequent are the IRMS value. This was the clue that let me find the problem. Reading the datasheet lead me to T9 the time you need to wait from writing the command register to reading the data. I had violated the minimum time. Adding a small delay (4usec) solved the whole problem.
So now I started to look at the accumulated energy readings and found that the active energy is decreasing and the apparent energy is increasing. Another layer in the onion to peel back.
On another note, the reboot issue with my file server has stopped as suddenly as it started. I have not had a single reboot now in nearly two weeks. What was all that about I wonder?