That's typical of people who "don't think" but, instead, sieze onto some "cliche" of techspeak and adopt it as a mantra: "avoid single points of failure", "don't use recursive algorithms", "don't run with scissors", etc.
Ask one of these guys what he's done to protect his network *fabric*! (which is usually more tedious to troubleshoot *and* repair than a server going down!)
Instead of thinking about the issue that is trying to be addressed with these little mantras, they take them literally.
E.g., I have *excellent* up-time here -- I keep cold spares of every bit of kit! Why should I run a name server *and* a time server *and* a print server *and* a firewall *and* a print server *and* a TFTP server
*and* a ... when I can solve *all* of those needs with a single host (that *still* has a loadavg of