What causes this resistor to fail - open circuit and no visible damage?

It is my experience (failure analysis) that 1% metal film resistors are not reliable long term with high voltage DC on them.

I've also seen huge changes in value without them complete going open.

You might try to fit a MOF (metal oxide film- also known as metal glaze) type if you have room. I recommend Vishay (BC Components nee Philips) VR25 resistors for this service if you care about reliability. They're tested at 100VDC, but have a "rating" of 1600VDC.

There's a hint here that moisture (probably plus elements of the lacquer) play a role, so if you want to recreate the failures I'd suggest putting them in your environmental chamber at elevated temperature and 95% RH and see what happens. Obviously, you want to avoid condensing humidity when power is applied or bad things (tm) could happen.

R10 is seeing almost all of the 220V input full wave rectified (there is a voltage divider with R12 but R10 is more than 100 times larger). The peak voltage across R10 is thus given by 220*1.414 which is 311 volts. This is larger than the specified working voltage of 300 volts. It is quite possible over time that this overvoltage is causing the resistor to fail. Since it is not due to overtemperature but overvoltage, the failure will not be visible on the outside. Most likely, there is internal failure of the insulation. In any event it would be advisable to get a resistor with a higher voltage rating.

The resistor in question is not protected against lightning strikes to live and neutral relative to earth. This could be the problem I.e. both L and N get raised to some absurdly high transient voltage whilst earth remains pretty much steady at ground potential.