Webots stuck in reinforcement learning simulations

[annaa95]

Hi, I've developed a reinforcement learning framework in Webots, starting from the deepbots library, with modifications to integrate a ROS controller (extern controller in ROS noetic).
I have a supervisor in charge of running the main RL algorithm that includes the resetting of the physics after each episode. A robotic agent is spawned at each episode and a ROS node is instantiated as well.
After a few epochs the simulation get stucked during the supervisor.simulationReset() call.
Any suggestion on how to prevent this?
I'm running Webots R2022a on a Ubuntu 20.04.