You're up against the laws of physics for a linear regulator, they'll make more heat with higher input voltage- regardless of any circuit mods.
The schematic shows 72V, you have 75V surely from mains being higher than 230VAC.
It's from an era where electrolytic capacitors were big and expensive so a very lean design. They also were conservatively rated, the old Spragues came +50% at least and could be run at rated voltage.
The only reason I could see for it to go "unstable" with lower input voltage would perhaps be mains ripple showing up and modulating the output.
What did you use for replacements on the filter caps, C2 80uF 75V, C10 65uF 60V +100%/-10%? You kept the same critical GND point "J" for the cap's (-)?
I find I have to round up as modern electrolytics are low value i.e. 82µF ±20% measure 70µF, whereas old electrolytics were on the other end i.e 95µF. I would double their value and then try it with lower mains input.
Otherwise, I would say it's meant to run hot.
https://www.eevblog.com/forum/testgear/hp-3410a-ac-microvoltmeter-restoration