strange routing behaviour with openvpn

[Neo [GC]]

Hi,

I'm using openvpn to connect my vpn-gateway at home to an external 
server, both are FreeBSD-boxes (6.2-STABLE).
The external server has an fixed IP, the client at home connects to a 
router, which gets a new IP every 24 hours.
The client is configured as router (gateway_enable="YES") which works... 
at least sometimes.

After my router gets a new IP, the openvpn client reconnects to the 
server and the tunnel is usable from FreeBSD at home to FreeBSD at external.
But: I have one Vista and one OSX at home, both have static routes to 
the FreeBSD-box. They are able to use the tunnel, when the 
openvpn-client is freshly startet. After getting a new IP from my ISP, 
the tunnel is up (and - as I wrote - the FreeBSD is able to use it), but 
the two other boxes don't get routed through the tunnel, but the default 
gateway. When I restart the openvpn-client, everythink works again like 
it should.

Sample output of traceroute when openvpn is restarted:
  1    <1 ms    <1 ms    <1 ms  wintermute [192.168.2.2]
  2    29 ms    30 ms    32 ms  GOTHNET [10.10.0.1]
(works)

After router gets net external IP:
  1    <1 ms    <1 ms    <1 ms  wintermute [192.168.2.2]
  2    <1 ms    <1 ms    <1 ms  192.168.2.1
  3    19 ms    19 ms    19 ms  217.0.119.195
  4  217.0.89.70  meldet: Zielhost nicht erreichbar.

Any advice? :(


Regards,
Neo [GC]