Hmm. I’m surprised you’re not having problems with Fume 5.0 as there’s a thread here with others having problems with it.
On to your actual issue here… I think both you and I expect “bad allocation” to be due to Stoke trying to allocate memory and the request failing. I’d expect it to push to virtual memory in that case too… Are you able to check the memory usage when you’re bumping the particle count up? I’m not sure about memory fragmentation either, but a reboot as a test would clear that up.