asterisk users - Jun 2007 - Polycoms lose registration and won't re-register

For the last few months we have intermittently been experiencing some very
strange registration problems with certain polycom phones.

Here is some background information:
I have about 150 Polycom Soundpoint IP 600s, 601s, and 650s spread between 8
servers at different locations.  Each phone is on the same network (and
subnet) as the server it connects to.  There is no NAT or anything else
strange that should be messing with the connection between the phones and
asterisk.  We are using Netgear FSM7328P and FSM7352 POE switches, with both
the server and phones directly connected to the same switch.  We have been
experiencing this issue intermittently for several months with all 3 types
of Polycom phones that we own. (600, 601, and 650)  It has happened with
phone firmware versions ranging from 1.6.6 to 2.1.1.  Asterisk versions have
been since at least 1.2.15, and it is still occurring with 1.2.18.

Seemingly randomly, a single phone will stop registering with asterisk.  All
of the other phones continue to work fine.  A "sip show peers" will
show the
non-registered extension as:

Name/username              Host            Dyn Nat ACL Port     Status
3310/3310                  (Unspecified)    D          0        UNKNOWN

The phone can still make outgoing calls, but any calls to it will go
straight to voicemail.

Rebooting the phone (by the keypad, or by removing power) will not cause it
to re-register, nor will stopping asterisk and restarting it.

This has happened on phones that use realtime, and on ones that are manually
set up in the sip.conf.

The phones are provisioned via ftp, and if I take a different phone than the
misbehaving one and rename the <macaddress>.cfg's from the misbehaving
phone
to the new phone, the new phone will always work fine.  The phone that
stopped working will continue not to register even if it is moved to a
different extension.  We have also tried 'touch'ing all of the config
files
for a phone that won't register in order to update the timestamps.  It did
not make a difference.

If the phone that refuses to register is moved to a completely different
location and server, it will begin working again fine.  It can then be moved
back to the original location/server and will be fine.

We always start with the stock sip.cfg/phone.cfg/etc for whatever firmware
version we are using, and then make a few very minor changes.

Below are excerpts from the registration section from the polycom
"phone3310.cfg" and "sip.cfg" for the 3310 extension I am
currently
fighting.  It is using the 2.1.1 firmware and connecting to asterisk 1.2.18.
These excerpts are from the normal config files we usually use, but I have
also tried changing the transport to "UDPOnly" and explicitly setting
the
registration expiration and overlap in the configs to shorter values than
the default of 3600/60.

Over the last few months, this has happened several times to about 7 phones
on 5 different servers.  Since I have a phone that is currently doing this,
I would be happy to capture any sort of debug output that may help determine
the cause of the problem.  Any help or suggestions would be greatly
appreciated!

Thanks, 
Eric

Excerpt from phone3310.cfg:
   <reg
      reg.1.displayName="3310"
      reg.1.address="3310"
      reg.1.label="Line"
      reg.1.type="private"
      reg.1.lcs=""
      reg.1.thirdPartyName=""
      reg.1.auth.userId="3310"
      reg.1.auth.password="3310"
      reg.1.server.1.address="192.168.35.1"
      reg.1.server.1.port="5060"
      reg.1.server.1.transport="DNSnaptr"
      reg.1.server.1.expires=""
      reg.1.server.1.expires.overlap=""
      reg.1.server.1.register="1"
      reg.1.server.1.retryTimeOut=""
      reg.1.server.1.retryMaxCount=""
      reg.1.server.1.expires.lineSeize=""
      reg.1.server.1.lcs=""
      reg.1.outboundProxy.address=""
      reg.1.outboundProxy.port=""
      reg.1.outboundProxy.transport=""
      reg.1.acd-login-logout="0"
      reg.1.acd-agent-available="0"
      reg.1.proxyRequire=""
      reg.1.ringType="2"
      reg.1.lineKeys="3"
      reg.1.callsPerLineKey="1"/>

Excerpt from sip.cfg:
      <server
         voIpProt.server.1.address="192.168.35.1"
         voIpProt.server.1.port="5060"
         voIpProt.server.1.transport="DNSnaptr"
         voIpProt.server.1.expires=""
         voIpProt.server.1.expires.overlap=""
         voIpProt.server.1.register="1"
         voIpProt.server.1.retryTimeOut="0"
         voIpProt.server.1.retryMaxCount="0"
         voIpProt.server.1.expires.lineSeize="30"
         voIpProt.server.1.lcs=""
         voIpProt.server.dhcp.available=""
         voIpProt.server.dhcp.option=""
         voIpProt.server.dhcp.type=""/>

I have made a little progress with this problem today, but am still looking
for suggestions as to what could be wrong:
 
>Rebooting the phone (by the keypad, or by removing power) will 
>not cause it to re-register, nor will stopping asterisk and 
>restarting it.
>
>If the phone that refuses to register is moved to a completely 
>different location and server, it will begin working again 
>fine.  It can then be moved back to the original 
>location/server and will be fine.
All of our phones use DHCP to get an IP address.  While playing with the
mis-behaving phone today, I forced our DHCP server to give the
non-registering phone a different IP address when it was rebooted.  After
the phone came back up with a different IP address it is working fine again.

I don't understand why that would make any difference, but it makes me think
that something is getting cached on either Asterisk or the Polycom phone
that is causing the problem.

Any help or suggestions would be greatly appreciated!

Thanks,
Eric