I’d agree that generally you shouldn’t send this gateway metadata via TTN.
A “real” gateway deployment should include a remote management solution anyway, and that’s probably a better place for your monitoring feed. Especially as you have a chance of using an independent feed of information to get an idea of the nature of the failure, if you stop getting any node data through TTN servers (ie, the servers might be the problem).
If you really wanted to make the gateway be a node, rather than messing around with a second radio you could just calculate a packet as if it had come from a send-only ABP node (at a fast spreading factor and moderate power) and inject it into the backhaul scheme. Arguably slightly abusive, but in no practical sense any moreso then having the box send data to itself by radio. In any practical sense, less abusive as it won’t consume any shared radio-related capacity. (If you were going to use a radio chip, have it feed a 50 ohm resistor instead of an antenna)
But you’re probably not going to find much willingness to help with doing this, as it looks too much like (and practically helps with) the misguided question posted about every other month of “how can I make the output of my Ethernet connected sensors look like it came from LoRaWAN nodes” and we definitely do not want people doing that.
So, report your gateway status through an independent channel, probably connected to whatever you chose to use for remote administration.