I enabled sofia tracing while the problem was occurring. I haven't had a chance to look into the source code yet, but so far I'm guessing it's not a permissions issue and something more like the invalid connection handle caching hypothesis. Here's the sofia log snippet (DNS names and IP addresses sanitized): nta.c:10803 outgoing_query_a() nta: for "myvoip.example.com" query "myvoip.example.com" A (cached) nta.c:10856 outgoing_answer_a() nta: myvoip.example.com. IN A 10.0.0.1 tport.c:3257 tport_tsend() tport_tsend(0xb5b392e8) tpn = udp/10.0.0.1:5060 tport.c:4046 tport_resolve() tport_resolve addrinfo = 10.0.0.1:5060 tport.c:4680 tport_by_addrinfo() tport_by_addrinfo(0xb5b392e8): not found by name udp/10.0.0.1:5060 tport.c:3636 tport_send_fatal() tport_vsend(0xb5b392e8): Invalid argument with (s=22 udp/10.0.0.1:5060) tport.c:3492 tport_send_msg() tport_vsend returned -1 David