Oh, I thought DNS seed was something along the lines of consult a main key domain such as for us devcoin.org or devcoin.com or devcoin.net and fine in its DNS somehow a whole list of IP addresses of nodes various people all over the net on gosh knows what domains of their own are running 24/7 as seed nodes.
With Round Robin DNS you can do that - but with ipv4 it's manual - so you create A records for node.devcoin.org for each IP.
ipv6 allows more of a dyamic lookup allowing more of dynamic approach - take a look at the srv records here - http://www.debianhelp.co.uk/dnsrecords.htm
Maybe it isn't really some special tricky use of DNS at all, maybe all they really mean is that if you have the ability to look up names to resolve IP addresses, then you can name nodes in the code instead of having to put their IP addresses?
Aye - you can plug in host-names instead of the ip addresses - giving you a little more flexibility with moving nodes about.The lvl of security is about the same - although I'd suggest adding extra logging for the name -> IP lookup - so you can debug issues should they arise later.
Risks..
About the same as now - to subvert you'd have to add a hosts entry - rather than currently (with ip) you'd have to add a route.
Resolving a dns name to an IP shouldn't cause a crash if you properly wrap the call.
Robust coding.. is robust
