Okay, this is going to sound crazy, as it did to me the first time the light bulb came on, but I can reproduce it very reliably now.
I have had the same Netgear WGT624v2 up and running since it was first available (several years, don't remember exactly when). I've had systems running Windows, several flavors of Linux and OS X (PPC) running on a network behind it during that time with no problems. I've used nics from Intel, Broadcom, Realtek, and a few others, both 10/100 and gigabit ethernet. I also have several gigabit ethernet switches (one large in a central wiring closet) and one smaller one used basically as an expansion hub in a home office. No problemo.
The OS X system I've had for almost 2 years is a Powermac G5, and it has never caused any problems either.
I bought a Macbook Pro (MBP), Intel-based, a while back, and not too long afterward (but I didn't make the connection right away) I started getting strange connectivity problems where the router would stop communicating (wireless and wired side both) on all internal systems.
Here's what I finally tracked it down to:1) It /only/ happens when I am using the MBP on the network. 2) IT does /not/ matter whether I use the wired nic port or the wireless on the MBP. 3) If the traffic is light on the MBP (like simple web browsing) then nothing bad ever happens. But, if I bash on it, like large file transfers (through to the outside internet) or browse a large site with a lot of large images or something, then the router will die such that it appears to be ok, but traffic won't be routed. Internal networking through it (LAN only) will still work. 4) The Powermac can not be induced into doing this no matter how hard I bash on it, same with the other PC-based boxes.
I can't blame it on a single network interface, since both wired and wireless do it on the MBP. It's not something in a particular file transfer causing the problem. The exact same transfer that will kill the router on the MBP will work fine on the Powermac or any PC system. I could be tempted to blame it on the router, but it works fine as long as I don't do any heavy network traffic on the MBP (months and months with no need for even a reboot of the router).
I'm wondering, since it effects both network interfaces, if it could be something slightly flaky with timing or protocol implementation on the Intel port of the tcp or udp code causing it? I tried a few variations on google searches for this, and can't come up with a similar problem so far. Also, firmware on the router is up to date.