The Segfaults that you (and most everyone else) sees regularly while trying to run OpenBSD as a guest are a result of incorrect cpu register handling and poor x86 emulation on behalf of VirtualBox. OpenBSD's strong security record is partially a result of the fundamental goal of correctness
, which means that OpenBSD is much less tolerant of flaky or out-of-spec hardware, whether that hardware is real or virtual.
Which is really unfortunate, because a thin, BSD-based operating system is especially good for things like embedded systems, and having a virtualized system to use for development and QA is a great thing.
The problem here is that the system that VirtualBox is emulating, is just not "real" enough to fool OpenBSD, particularly if VT-x is not turned on or not supported by the host processor. Hopefully, the 2.0 rewrite that Sun is doing will help to fix this.
I have had limited success using the -norawr0 option to VBoxHeadless and VBoxSDL (this is sufficient to get OpenBSD 4.3 to install on my non-VT-x hardware), but OpenBSD still throws random segfaults and other nasty behavior, when put under any sort of load (try rebuilding the system, for instance) -- in practice, it is pretty unusable... and it's completely useless for any sort of QA or other testing work.
I haven't really put NetBSD through its paces on either of my vbox builds yet (Sun's non-free one, and a fresh SVN build), but I'm not optimistic -- it also requires the -norawr0 option to even install.