I would probably recommend something like this instead of the below change; in particular, it defends against the additional case below where we try to concatenate nil in the citype ~= "postnet" case. OTOH, I'm looking at that and content probably suffers from the same problem. I wondered if it was better to move the new condition up here, but I guess we probably still want to write the blank output to /var/cache/nuageinit?

OK, yes, it is better.
However, I'm not sure if we want /var/cache/nuageinit with just a newline?

Right, maybe not- if it doesn't make sense to preserve an empty user_data, then you might as well preserve the nil-ness (so that you know it's empty and not just a blank line) and just move the check up. I wasn't really sure if it's feasible to reach this on a non-first boot and need to do something still with the user_data file written, but I guess the below open doesn't truncate it (and there's definitely no value in adding an extra blank line to it).

This checks that the first line is not blank (no white spaces, tabs, non-printable characters, etc.).

If you think this blank-string check could be useful in other parts of nuageinit, I can create a function in nuage.lua.