iterative query timeouts & recursive-clients

Hi,

Just wondering; when a caching server (Bind9) sends out queries, as part
of the iterative resolving process, how long does it take for the query to
time-out? Can this be tuned or is it hardcoded? Is it specified (in e.g.
some RFC) how 'patient' a server sould be in this case? Or based on ??

Also I'm wondering what 'recursive-clients' clients option does exactly.
If two clients request the same name (& record type) at (roughly) the same
time and the answer is not in the cache. Does this count as one, or as two
towards the recursive-clients limit? It only needs to do all the queries
*once* in order to answer *both* clients, but I'm uncertain how to
interpret "The maximum number of simultaneous recursive lookups the server
will perform on behalf of clients." in this case.

Coupled with the first question; How quickly can the number of
'recursive-clients' be exhausted when some (or all) name-servers needed to
resolve some freq. requsted name are unreachable?

Background. Just some questions that where raised when thinking about the
"no more recursive clients" log messages. recursive-clients is currently
set at 10000 and still (sometimes) beeing hit... And I'm thinking about
which things I may need to take into account when guestimating what are
reasonable values for that option.

Rgds,
Mark.