CUPS.org User: mike
EAI_AGAIN is a non-fatal error and should just be retried (which I'm pretty sure we do, at least in the backends...) Reinitializing the whole stack is really the wrong thing to do.
EAI_NONAME errors are not cached, or at least they aren't supposed to be. If they are, that is a glibc bug.
Really don't want to take this patch, sorry...
diff -up cups-1.3.9/cups/http-addrlist.c.res_init cups-1.3.9/cups/http-addrlist.c