I used to work almost exclusively on mainframe UNIX servers... downtime was NOT allowed - except at a specified monthly preventative maintenance period. And those periods could be removed at the drop of a hat. Desktop systems were treated pretty much the same - they would get even less maintenance - and still had to be up and running for months at a time. UNIX could do that, Linux as well...

tmpfs has its uses - the first use (though it wasn't called tmpfs) was for file locking. Using ram for a small filesystem was MUCH faster for handling and recovering locks. A system reboot always restored the locks to a known state.

Not sure why this is the case, but 'systemctl unmask tmp.mount' - which left tmp.mount disabled - allowed my fstab entry to work. This is on CentOS 7. (I, too, and working on a mostly CIS-compliant install).

The only thing I can think of is that systemd was using some stored state and not regenerating it when the fstab entry was added. Mask/reboot/unmask may have forced it.