It appears that there was a slight degradation in performance again (from around x3000 to x2600), which suggest that some unrelated stuff is being calculated when it doesn't need to be.
I'm sending by email a Firefox profiler capture of around 1000y of fights in the wilderness together with the related save, hopefully that helps with finding out what the bottleneck is, unfortunately I couldn't parse it much myself.
Note that the profiler caused an additional overhead, which brought it down to x2000 during the recording.
One thing that is curious is that around the 10sec marker, which is around when the performance starts to degrade, the Parent Process memory is unloaded; The most likely problem that I can see is the DOM Worker getting overwhelmed by repeated calls during high speeds, but I have honestly no real idea how to optimize something like that.