Power into the lamp is proportional to the voltage non squared due to the arc voltage being constant. Only the current changes and that is linear with the v drop on the ballast
What matter with magnetic ballasts, is the ballast losses. And because the ballasts are more-less linear devices, their losses are proportional to square of current, so square of the voltage.
It is mainly the ballasts, what make the lamp circuit so vulnerable towards long time, but even slight mains overvoltages. Extra 10% of current mean only 10% lamp power increase, but ~20% ballast losses increase. And the extra 20% of ballast losses mean their temperature rise would be 20% higher.
In order to keep the mass reasonable, the ballast design have very little margin for any extra dissipated power: The designed temperature rise is usually about 70degC, already very close to the limit for usable life.
Therefore the long term overvoltage shall never exceed 5%, what mean about 6..7% current increase in higher arc voltage circuits (like 40/36W, lower arc voltage circuitss suffer less, as the ballast voltage is already high), what is about 12..15% power dissipation increase, so the temperature rise would become nearly 80degC. With 40degC ambient it mean 120degC in the winding...
With 10% voltage rise the current rise by about 12..15% (still not big deal for lamps), so ballast power dissipation increase by 25..30%, so temperature rise about 90..100degC. That mean the winding could reach easily more than 130degC, what could be survived only for shorter time.
Of course, the thermal budget strongly depend on fixture's thermal properties, the figures belong to more-less worst case scenarios...
But don't forget the winding is not the only thing affected by the elevated temperature. All the "crispy plastic" connector issues are strongly linked to the ballast operating temperature as well, so although the ballast itself may survive (they are made quite tough), the surrounding components could fail solely due to the elevated ballast temperature...