Having written (as in, programmed, based on process knowledge from fire departments) a model for forest fires in Australia myself over the last few months, I feel that I can say that while a fine demo application of CA based modeling, this model is not an accurate description of forest fir behaviour (well unless it's just to illustrate how planned burns interact with their crude approximation of fuel load). E.g. it doesn't even consider slope and wind direction, or vegetation type.

Some places getting dramatically hotter drier weather all the time does not lead to less severe fires.

I know that some trees have adapted to fires. Some pine cones are held closed by a resin that is sensitive to high temperatures. These cones will not open to release their seeds until the critical temperature is reached, out with the old and in with the new and such.

However, if there's nothing left ot burn and the whole ecosystem is destroyed, that's a problem

Not paradoxically. No tree burns two times, so in the next fire you'll have less burning material. This is obvious, and bad.