On Erlang 20 long GCs run on the thread pool for “dirty” (blocking) tasks, they don’t block a scheduler, so the real-time properties are nice. There have been

On Erlang 20 long GCs run on the thread pool for “dirty” (blocking) tasks, they don’t block a scheduler, so the real-time properties are nice. There have been implementations of Erlang with shared heap in the past, before BEAM became heavily multi-threaded: https://www.researchgate.net/publication/2371933_A_Case_for_the_Unified_Heap_Approach_to_Erlang_Memory_Management https://www.researchgate.net/publication/2371933_A_Case_for_... I think it would be nice to see this tried again. Perhaps the issue is scaling to future systems with potentially hundreds of cores?