I'd recommend 1962 RCA Electron Tube Design book - it has an example thermal assessment based on 6L6 output stage tube. Due to the T^4 - T^4 radiative transfer relationship the anode temp doesn't reduce much for a significant drop in glass temp, so there is not much net increase in anode temp obtainable to give a higher anode dissipation capability. Forced air convection is the most practical, and seperating the valve from its neighbours (and not aligning the hotter portion of the anode to the next valves hotter portion) are the typical practical aims. The main benefit from better cooling of the glass is reduced outgassing (and hence lifetime of the getter material). Pushing an anode to a higher temp than nominal has the same detriment to outgassing. It's probably not a great idea to reduce heater power to then allow higher anode dissipation - for a net zero change in total internal dissipation - as the cathode chemistry may suffer from too low a temp.
Is your question just for fun, or is there an actual target application?
just for curiosity, always pondered about pushing a tube to it's absolute limits in an amplifier, even built some class B2 EL34 amps with cathode followers driving the control grids...
anyway, I tested it with a random old 6P15P tube I had (RF version of 6P14P/EL84), because it was the only tube I wouldn't be sad about destroying.
in the outside air it would handle 14watts (that's anode + screen dissipation) just before starting to glow on the anode, and in a jar of water and stirring it constantly I could achieve 15watts.
heater voltage was 6.9volts (due to 220V transformer it was high)
the screen was supplied from the anode by a 5.6k resistor, anode voltage was around 230V
used a 200ohm cathode resistor with a pot on the control grid adjusting between ground and cathode voltage.
so it seems kinda worthless, a chimney around the tube would probably achieve the similar gains anyway.
maybe I'll test dropping the heater voltage now.
speaking of that, I found a chart for the 6P15P coincidentally that shows change in plate current, screen current, and transconductance relative to heater voltage.
found it on a thread about heater voltage, apparently many tube amplifiers ran the preamp tubes (12AX7s) at reduced heater voltage ~85% lower for increased linearity.
and some other tubes like 6AL5, and 12AV6 have better linearity at reduced heater voltage.
here's some charts on that: