Problem with Secondary systems with many zones

[list3 at wwwcrazy.com]

This problem that I have is with systems that have many zones (at least 30,000
zones).

If I start a secondary system up with no zone files (so it is completely clean)
the system will eventually get all of the zone files by getting an AXFR from
the primary name server.  Then once it gets all of the zones it works perfect
(dynamic updates and the whole bit).

When I start a secondary system with zone file already in place (often after a
reboot) the sytem will never finish the "soa queries in progress".  They just
keep climbing up until they hit a high number (close to the number of zones in
the system) and the system just sits there.  It responds to queries but it will
never get any dynamic updates nor will it ever refresh the zones.

number of zones: 62786
debug level: 0
xfers running: 0
xfers deferred: 0
soa queries in progress: 62780
query logging is OFF
server is up and running

So my solution has been to stop bind.  Remove all of the zone files, and start
again.

The zone files are in a hash diretory structure.

so example.com would be in "e/x/db.example.com"

Any suggestions on why the secondaries on Bind seem not to be able to handle a
restart when it has many zones.
I have other sytems that have 15,000 zones and there is no problem.
This only happens with about 40,000 zones and higher.

Thanks for any suggestions.
-Steve