Do you have a tach and really know what RPMs you are turning? I use
this one. I found early on that 10-15 perforations per millimeter of travel seemed to give decent cuts for DTFB (paper on)... so came up with a "rule of 10"-type ballpark estimate for feed and speed; i.e. 600 mm/min and 6000 rpm or similar. That works out to 10 perfs/mm and I could stretch it to 600 and 9000 for 15 perfs/mm. Above that, I would start seeing the black specs and needle coating with my cutters at the time. It appears that for 10000/480 you are running in the 20+ perfs/mm range.
I think what is happening is that the increased cutter RPMs result in increased Mig-tip friction heating that transfers to the steel needle... which results in highly localized melting of the foam (hard black specs/"gnat droppings" and shiny/glassy-looking foam edge in my photos), which clings to the needle, and eventually leads to failure of some sort; i.e. flying/broken needle, clog/seized needle guide, etc.
If your cutter otherwise runs comfortably at the speeds you've been using, I'd suggest increasing your feed rate to lower the amount of time the needle spends in an area... say, 900-1000 mm/min (15-17 mm/s) for 10-11 perfs/mm. The parts won't generally fall out on their own but should still easily pop out of the sheet. I found concentrating on 10-15 perfs/mm made it easier to find a feed/speed "sweet spot" that still resulted in decent cuts. You might also try slowing the cutter RPMs and/or remove the Mig-tip heat with some form of active/passive cooling. Also recognize the needle cutter is an inherently "sloppy" mechanism by necessity and will probably last longer the less it's "stressed"... i.e. absolute fewest RPMs necessary for decent cuts in acceptable time frame. Just sayin'...
Good luck! -- David