Previous runtimes are not stopped and they live alongside new runtime. When we detect hot-reload we want to: 1) start new runtime, 2) stop old runtime.
The order is intentional. We want to replace server handlers after new environment is ready. However, this may lead to problems when there are layers that operate on the same resource and require a global lock leading to timing issues. This setup is very unlikely in development setup though.
Previous runtimes are not stopped and they live alongside new runtime. When we detect hot-reload we want to: 1) start new runtime, 2) stop old runtime.
The order is intentional. We want to replace server handlers after new environment is ready. However, this may lead to problems when there are layers that operate on the same resource and require a global lock leading to timing issues. This setup is very unlikely in development setup though.