The spec sheet for a high voltage surface mount resistor (Bourns CHV2512) says the maximum working voltage is 3000V; but the nominal lead spacing is 182 mils. The on line PCB Voltage Spacing calculator, says the maximum voltage for uncoated conductors would be 860V and about twice that with conformal coating (but of course the land pattern normally is less than lead spacing).
What gives? break down voltage in air is about 1KV/mm, so 182 mils = 4.6mm or about 4.6KV. Is the PCB Voltage spacing calculator extremely conservative and/or the spec sheet overstated?
I need about 5KV across 5M ohms (nominally the current is 1 uA, but in the event of arcing a 2.8nF capacitor at 5KV is discharged. ).
What is best practice here?
With a conformal coating is it reasonable to assume I would need 4 or 5 resistors in series?
Firstly you have done a great analysis of the resistor. and you will have to use not as much as 4 - 5 but at least 2 - 3 resistors in series for the 5kV to be divided. nominally you will have not more than 5V across the resistors with 1 uA of nominal current.
The PCB voltage and spacing calculator tool is derived from the IPC2221B table and the values are a bit conservative. The specs sheet of the resistor will definitely not be overstated and you can follow the data provided in the datasheet.
If you are still concerned you can reduce the land pattern size by 10mils keeping the length and the width the same. that will increase the distance between the land pads from 183 mils to about 200-203 mils, with no issue while assembly.