I have looked into this a bit, and I'm assuming the problem is the same as this guy?
Edit: After digging a little deeper, I have also stumbled on this 6 year old thread when Smiff was still around looking for input for the official RDB updates.
Post #415, user "Thunderbird8" states: Not sure if this is all that helpful, but Rayman 2 seems to need a high Counter Factor in the Cave of Bad Dreams. If it's too low, the cave master will miss one of his jumps and fall into the pool of...stuff...below. A CF of 5 worked, but I skipped 4 when testing. I haven't seen any problems up to this point with a CF of 2 though, other than the aforementioned issue.
I am curious on how this change affects performance and stability in other parts of the game. If this is a viable fix and does not affect performance, it could very well be set once and never need modified again. Default PJ64 plugins were supposed to be used for these tests, I am guessing he experienced this issue with Jabo's. Since you got past this boss using a more recent version of the plugin, the question remains whether or not it's a timing issue with emulators or clipping issues with graphics plugins. It may be possible you got lucky and it happened to work the time you did use Jabo's. I am leaning more towards a timing issue. It would also be interesting to know how many emulators or plugins suffer this issue. This will require some experimentation.
This issue is also reported in the Mupen64plus compatibility list.
If you happen to have a save state or SRAM before this boss, can you upload it please?