The emergency lights are operate the lamps just for few hours, so the lamp munching is not that big problem for that application.
For the battery lanterns, the thing starts from the total cost calculation:
The most expensive consumable are the batteries, so the design is made so to not eat them excessively for the desired light output specs. So the efficacy is the top priority.
You may reach the light specs by either using lower wattage tube and drive it at its full rating, so the tube will last long there. Or other option is to use larger tube, which uses to be way more efficient and drive it at low power. This means the tube may suffer, so needs to be replaced after 100 hours or so, but you are able to get higher efficacy, so lower battery drain. So although you are changing tubes rather regularly, they still cost way less than would be the extra battery cost with the lower power tube. Because running the tubes DC means just one electrode needs to be heated, so most of them run the tubes by only one polarity (AC feeding with the same power and lifetime would need both electrodes heated, so double power needed for that; the cathode does not need external power, it could be heated just by the ion bombardment alone, in fact that is the most energy efficient method, although it means a bit shorter life).
This is the main reason, why the battery lanterns are all designed to use 6W tubes at 2..3W power and 8W tubes at 3..4W, it is the sweet spot of the lowest power for the given light output.
Even the cathode heating when used, powers just one cathode. The lamp may be connected by 4 wires, but the connection on the anode side usually serve as missing lamp cutout and do not provide any heating power (anode does not need heating at all anyway).
So regarding the upgrade, it depends what you want to achieve.
The lowest operating cost? Then leave the original circuit alone.
Better reliability? That is tricky. You may add some circuitry to prevent excessive load on components during abnormal operation (e.g. failing lamp,...), but that would mean the lantern would more often go into shut down mode, so in fact become less reliable as a light source at the moment when it is needed. As leaving you at dark is considered worse than this being the last few hours of the lanterns life. So by not implementing the shut down you get longer time with at least some light in the wild (when e.g. attempting to fix the broken car).
The only improvement what came to my mind as somehow useful is to make the ballast less dependent on the battery voltage, so do not drive the lamp at that high power with fresh batteries, but then better maintain that chosen power when they are dying. Even this is a two edged sword: It allows longer time operation, but it will exhibit way sharper decline when the batteries are about to die for good. The standard behavior means based on the lamp performance you immediately see how long the batteries are going to last, so it has its beauty too.
This latest mod is very simple: The ballasts are usually some form of a blocking oscillator, with some resistor from the supply to the transistor base. This resistor dictates the peak collector current, as well as it influences the off time during the operation. Generally most inverter circuits have their output power directly proportional to the current through this resistor. So by making this current constant over the battrey voltage variation you make the lamp power constant as well and viola, you are done.
In
this circuit the R3 is the resistor in question. You can see the heater winding connected to just the cathode side, the missing lamp interlock is not implemented there.
So in short in battery lanterns with very limited power the requirements are very different from the lanterns where the available power is way higher (mains, motor vehicle illuminations,...).