Hi!
I'm trapped into same situation: FreeBSD on zfs inside Virtualbox. Nothing worthy. But the fact is surpising for me! At first that zfs can die without any hope to recovery (I not use a lot of setups, so my skills are poor, yes all my setups without duplication), and at second that there can be hardware I/O disk errors only inside Virtualbox!
Yep, I have always believe that inside virtualbox we will never see any hard drive errors like on real hw, but YES, they can be!, as someone proposed on first page. After hang up of the host while running virtualbox guest I got unbootable guest system with the same (yes, unrecoverable in my case, as in yours) error. And when I do check the reliability (with dd) of the drive (guest drive), I've surprisingly got I/O errors, not only on FreeBSD, but with linux guest also. On the host system file can be read without any errors as file (also dd for example), but copying virtual drive by Virtualbos storage management tools fails.
May be it be useful: virtualbox have no instruments to check consistency of VDI, i reveal error only by copying/converting process. But qemu-img convert this corrupted VDI file to another (i choose VMDK) without any errors, and inside quest I at least see no hardware I/O errors (also, for example, using dd). But zfs pool still cannot be recovered....