How is PID 1 made special and unkillable?

Using docker, I recently found that PID 1 is by default unkillable by SIGTERM. Yet you can still catch SIGTERM in PID 1 if you setup a handler in the process. So this doesn't seem to be the case of the kernel just masking SIGTERM for PID 1. How is this implemented? How does the process as PID 1 have the default SIGTERM handler be a noop?