- Core starts N javascript VMs in a global pool to maintain availability
- Each VM gets populated with GCRuntime and other modules
- Each GDPS gets assigned with a smaller pool of VMs from the global pool every request
- Modules are loaded from cache separately in different runtimes
- Each runtime gets a timeout of 30s (adjustable)
- After there are no more js and messaging events, VMs get "cleaned" and returned to the global pool
- If timeout is reached, VMs are killed
See Runtime Readme