I did a simple measurement that maybe someone with better secondary test equipment can confirm.
I used the Siglent SDM3045X to probe my bench power supply, only in the positive lead I inserted a 1Mohm resistor.
I used my VC8145 to measure the voltage across that resistor.
With the Siglent in 6V Manual mode I started the PSU at 1V then gradually increased it.
At the point of the relay clicking the voltage read by the Siglent was 2.2V and just prior to the click the mV reading on the VC8145 was 400mV. Bearing in mind there are some difficult high impedances interacting here including that of the VC8145 in paralell with the 1Mohm, this equates to a current just prior to relay clicking of 4x10-7amps. Since the Siglent was reading 2.19volts at the time then with 4x10-7 flowing through it the resistance of the Siglent is 5.5M ohms.
Just after the relay clicked the voltage of the Siglent shot up to 2.89V but the voltage on the 1M resistor went down to 200mV. Current now therefore is 2x10-7 and input resistance of the Siglent, now measuring 2.89V, would appear to be 14.45 M ohm.
Whatver that relay is doing, and whatever the real input DC resistances are (I accept these measurements are very dodgy) there seems little doubt that the DC input resistance jumps around somewhat on the manual 6V range. Personally I like a meter on a manual range to do no switching whilst the value measured is within its range. I can't actually provide a circumstance where its likely to mess up a reading for me, but if there are times when DUT loading by the instrument may be important to some, and I would have thought it unhelpful for a manual range to flip around like this.
Incidentally, on going in the reverse order from higher volts downwards, the relay doesn't click until 2.00v, as measured on the Siglent, is reached so there is obviously some hysteresis in there - of course if there wasn't , the meter would go wild if measuring a voltage of exactly 2.2V
Incidentally on the downward slope the Siglent dropped instantly from 2.0 to 1.87 volts. Again showing that since the 1Mohm loading resistor clearly didn't change, then something inside the Siglent must have done.
Here's a challenge - does any other Bench Meter on the planet do this when set to a manual range (excepting when a range is exceeded of course, which isn't the case here) ?
Eloso