It can be displayed in a debug node, and can be read from there.
I think that is because it isn't a real environment variable but rather a node-red construct. Where Node-RED allows you to access it using the same node-red mechanism.
process.env
on the other hand does read the OS's environment variables out of system memory.
Though it probably only knows about the ones that existed when the current process started? Not sure, I haven't checked that but it would make sense.
That is the key benefit of using an environment variable, of whatever type.
I suppose that's a slight benefit, at least if you have read-only dashboard users who can't deploy flow changes.