I understand that IP addresses are not reliable in pin-pointing a person's
physical location. The inaccuracy is attributed to multiple factors. The
IP could be from the DHCP pool and shared between many devices. The IP could be a proxy.
So far in the early part of the book, the IP has already been used twice
to identify the physical location of the computer user.
I refuse to believe the IP was employed as a shortcut.
Stephenson does his research. So I'm not ready to accept it was accidental
or necessary just to move the plot forward.
Some recent research was able to pin down physical locations to within half a mile or so from IP addresses, using ping times from a lot of servers in known locations. So yes, the location finding in Reamde is plausible.
So far (I'm about the third through the book), it has been realistic. Blocks of IP addresses are allocated to an ISP, and that ISP would then break that down further. Each Wi-Fi network would need an address range, so such a network could be identified, and a Wi-Fi network naturally covers a relatively small area.
You can use an IP address to find where a package came from. Tracemyip.org can show you that your general location can be discovered. Its not a far stretch to assume a more exact location can be triangulated
Purchasing a "Static" IP is always possible and is in fact frequently done. The ISP which provides this static address would have the physical location of this endpoint. While the ISP is not likely to freely provide the location of this endpoint it is not unreasonable to presume a sufficiently talented (read hackerish) admin is able to attain this information.