Good diffusion data requires good temperature control and good stability. Run the self-tune on the VT unit to make sure you have optimum temperature stability. Turning VT off is only acceptable if you are satisfied with diffusion data at an unknown ambient temperature. Also, I wouldn't spin, as that might create vortex movement (like convection) in the sample.

You also need good lock stability, so make sure the environment around the system is stable during your measurements. Run the "loopadj" command to optimize the lock response, and make sure that the lock phase is correct and that you are not saturating the lock.

The water chemical shift depends on temperature, HDO signas may be affected by convection + sample mixing during the rotation. Try to turn heater and VT gas off and measure DOSY again.