During a field study this year I had our CR1000X set up to read 10 single-ended voltage measurments from 10 attached moisture sensors. The measurements were made by first powering all 10 units on with the switched 12 volt supply. Then a pause for 10 ms for the sensors to settle. Finally the single ended voltage measruements were made in sucession from channel 1 to channel 10.
Most of the deployments had a sensor attached to each of the ten channels. However, one test deployment had six sensors attached, and the last four channels were left open. For that test deployment, I didn't change the number of channels being read. I just measured all ten channels and knew I could ignore the last four.
As I have been looking at that test deployment data, I noticed that the four open channels measured a voltage signal that looks like the six channels that had sensors on them. If you plot a time series of all 10 channels, it isn't immediatly clear which channels had sensors attached and which ones were left open. I was more expecting noise on the four open channels. Where would those four open channels be getting a signal from?
What you are seeing is likely because the AtoD converter in the CR1000X is multiplexed. Each channel is internally switched to the measurment circuitry. The inputs are very high impedence. So, charge injected onto the analog circuitry from channel 6 barely gets drained reading channels 7 to 10.
I would guess that you are doing the 10 measurements as repetitions on a single VoltSE instruction. The internal configuration steps of draining voltage off the circuitry are not done between repetitions.