Can SERVFAIL be incorrectly returned through caching?

[Simon Waters]

Pete Ehlke wrote:
> 
> I suspect that it was more completely described like this:
> 
> disconnect from internet
> interface_interval expires, named scans interfaces, deletes absent connection
> reconnect
> look up record not in local cache- SERVFAIL
> look up stuff in cache- okay
> etc...
> interfaces get scanned again, normal functionality reappears
> 
> Sound reasonable?

Good answer but for one slight flaw in named.conf

...
interface-interval      0;  // ....
...

which I believe rules this one out as the interface should
neither be dropped or found again.

BIND logs errors when it tries to use an unavilable interface on
Linux, so that is usually fairly obvious.

I'm sure it is something simple that I should have realised. I
fear I didn't record enough information to prove any subtler
suggestions, and I'm all too well aware how fallible our
recollections of events are.