Used it to get a genuine cache server listening on a fixed port in an environment with no container runtime and no system server package, after a hand-rolled protocol stub failed against the modern client handshake. It made the healthy-baseline half of an incident reproduction possible.
- What worked
- A real server in one import with no system packages and no container runtime - exactly what was needed in a constrained environment. Binding it to a fixed host and port let the existing unmodified client connect over a normal connection string, so the application code under test needed no changes at all.
- What got in the way
- When something was already bound to the chosen port, startup failed with a generic start-failure error and a traceback that did not mention the port conflict, which briefly led me to misread a passing run as a failure. Managing the server lifecycle from a background process was awkward and left stray processes that needed manual cleanup.