It doesn't matter on this project since everyone gets a 100 score for network...

Blake's UD Tweaker is much more dangerous for WU being correct than UDMon, since the UD agent is not designed to & is designed not to have more than one instance at a time on a computer.
Before using UDMon, I could crunch only ~6 hours a day on average. Now I'm over 14, and I've recently added another computer, which is also running UDMon (although it currently has only one cache slot). As the others said, as long as it is passed the -wcg script, there's hardly any trouble. UDMon has been stable for quite some time now.

One thing to watch out for on dial-up connections is your modem connection speed. If your modem happens to negotiate a connection below 28.8kbps, then your machine will not meet the minimum requirements specification (which is officially listed as 40kbps, but we give a little margin) and won't download new work until it happens to try again when you have esablished a faster modem connection.

I have a v92 modem too, but once in a while it connects below 28.8 kbps perhaps because of some intermittent noise on the line or even a problem at the provider's end. So, it can happen.

As for ping, it uses UDP packets (instead of TCP) which are not guaranteed to arrive, so when the network is busy, these can be lost sometimes. Given that you do see some losses of UDP packets, it might even be the case that the work unit downloads are having difficulty getting completed before timing out.

The vendor software does not account for that very large amount of memory. Most projects would not use anywhere near that much, otherwise we would have very few members which would qualify to run it. The current Human Proteome Folding project uses about 200MB of virtual memory and only about 25MB for the working set size.

maybe WCG could consider creating an in-built version of WCGmon to the next version?

just an idea