That definitely looks like a problem. I think what is happening is that soilwat is saying: "You have that much N in the profile? I'm going to leach this much away.". SoilN is told the leached amount, but says "I don't care what you want, soilwat, I'm not going to let N drop below 1 ppm". This will happen whenever there's a heavy rain/irrigation event comes along, and there's more leaching. It's really a mass balance problem - soilwat thinks it's removed the N from the system although soiln doesn't actually allow its removal.
It would be interesting to know why this wasn't a problem in 3.6, but that's going back to some pretty old code.
It may be useful if someone would attach the zip archive with the sample simulation referred to in the pdf document.