This has been discussed before, but I would like to get some developer opinions on the issue. I'm not advocating for or against it, since there are of course ups and downs to a 64-bit build; it's definitely not a magic bullet. In my own experiences, 64bit builds of certain software can reveal previously undiscovered bugs. Also, a 64-bit firewall in particular has additional complications. Namely, that the additional overhead tends to result in more cache misses, which can cause packets to drop on the floor at higher volumes. For most environments, this will cause more harm than good.
On the other hand, there is one key advantage: 64-bit address space. Right now, even with PAE, kernel memory is limited to 32-bit address space. Therefore, a pfSense firewall can support somewhat less than 4 million simultaneous states. In certain environments, this disqualifies pfSense where it could otherwise be quite effective. This may be completely moot, since PF itself is by nature a single threaded beast and will likely be the bottleneck first and foremost. Due to the generally locky nature of OpenBSD I'm not sure that this will ever change. But at least the state table issue might be remedied... thoughts?