If the problem starts to occur a few days after the gateway has been rebooted, then the suspect would be the gateway is getting busy and is unable to route the packets in the appropriate amount of time. What do the system performance tools indicate? Does the processor tend to get busy after a few days?
I'd still be suspect of the Internet. Traceroutes and ping tests do not make for sufficient tests for VOIP traffic. Hook something else up to the network and try watching some remote video or audio that last for 20+ minutes. Are they consistent? Also, a 1.5MB up and 10MB down has very little to do with what you get. There are two common issues, first, they can provision the circuit for most anything, but that doesn't mean that the circuit can support it. It is common to buy one thing and get delivered another. And while you might have bought a 5 ft wide water pipe to connect to your water provider, if they only have a two inch line in front of your house, you are only going to get two inches worth of water (shared with everyone else on the feeder). Cable companies are notorious for having a lot of people (1000) on the same feeder, each promised 1.5 MB of up, but yet the cable company only would have maybe 45MB of connectivity to the Internet. Try rebooting the server on a daily basis and see if the problem occurs.
