The bad news is that the quality of the DNS content sucks.  The good
news is that it is sufficiently resiliant that it seems to not bother
folk much.  At least not sufficiently that anyone is doing much about it.

The quality is worse in parts of the namespace where utilization is light
and/or the cluefulness of the users is light so they don't know what has
bitten them.

I had an experience where bad weather took critical servers down for three
and a half days.  We learned

  o essentially, the protocols may be resiliant, but they suck for error
    checking when things scale up to the size they are now.

  o lame delegation reports miss far more than they catch, as they are
    haphazard.  

  o a simple audit found 236 lame forward zones pointing to one of my
    servers, and a similar number of reverses.  and many of these were quite
    old.

  o when a prim/secondary server is down for a long period, the other
    server going down for 15 minutes becomes very noticable

Some things you can do now:

  o validate your site's data
 
  o use reasonable secondary techniques, check out KRE's

    draft-ietf-dnsind-2ndry-00.txt

  o the NICs should tighten procedures

  o until the protocols and procedures are seriously improved, the net
    police should patrol the neighborhood and we need some block homes to
    try and help the neighbors

  o get the IETF to work on the architecture of the protocols and not
    indulge in bit fiddling, bickering, and junior demagogery

-30-