There is a very good reason. Systems slow to a crawl for a long period of time when Windows increases the swap file on the fly, and a static swap files ensures a user will hit a hard limit, eventually. If someone is running out RAM and swap, the hardware and software requirements need to be re-evaluated. The automatic expansion could masking the problem, and the user could be trying to be nice and continue to live with the poor performance. Basically, the system to complain, and in turn the user, rather then mask the problem.
It also makes monitoring easier if the swap space is static. 25% of a known 6GB is more meaningful then 25% of ?. (?=whatever the OS felt like setting the swap file to at the moment)
Vista started setting the swap file to a much more reasonable number then past versions, and that has negated this piece of advice somewhat. In this scenario, it's a troubleshooting measure. If the system is swapping heavily and the OS is masking the problem, then causing it to run out of swap space will narrow down the problem. If the system doesn't fail, then something is hogging the disk IO.
Linux has some more leeway vs Windows with the paging design through the swappiness value, but one of the kernel developers is apparently a firm believer in an aggressive swap design.Link
The Linux 2.6 kernel added a new kernel parameter called swappiness to let administrators tweak the way Linux swaps. It is a number from 0 to 100. In essence, higher values lead to more pages being swapped, and lower values lead to more applications being kept in memory, even if they are idle. Kernel maintainer Andrew Morton has said that he runs his desktop machines with a swappiness of 100, stating that "My point is that decreasing the tendency of the kernel to swap stuff out is wrong. You really don't want hundreds of megabytes of BloatyApp's untouched memory floating about in the machine. Get it out on the disk, use the memory for something useful."
It really depends on what you're trying to do. Setting the VMM to swap out aggressively will provide more usable RAM at the expense of responsiveness. If the user or computer is going to be doing one thing for a while, or be doing one resource intensive thing, it would be a good idea to setup an aggressive swap setting. Responsiveness requires less lag, so setting a less aggressive swap is better. Andrew is probably compiling lots of code and using an IDE the most, so he is a good candidate for a very aggressive swap setup.
I'm personally think a system should be as responsive as possible. RAM should be leaned on hard since it's going to be the most responsive storage medium, and I want to switch quickly between multiple applications. As such, I'm going to favor a less aggressive swap setting like 20 - 30.