This is actually mostly close to a year old by now, when vjek and myself had a conversation regarding some of the unk flags in entities. However, it wasn't recorded or progressed further, and while I haven't really made much progress, I though it might be useful to note a few things that we observed before it's completely forgotten (I may well miss some of it here):

• unk6: A good predictor of whether a civ is considered dead (by DF) or not. However, it can err both ways, so it probably really represent something else that correlates strongly with death.
• unk11: If a verified dead civ with unk6 and unreliable_lost_last_site being set has both unk11 and unreliable_lost_last_site cleared, the civ appears as a live one (well, struggling) when embarking, with an immediate monarch appointed with the non standard message "X has claimed the position of king". My current experiment has the dead civ build up a reasonable pop (close to 100, including other races), retiring, flipping the flags, and embarking. As far as I can see it works as a normal live civ (positions visible on embark, including a monarch (human queen at the old site, one of the citizens there), dwarven caravans, trade liaisons, and 3 migrant waves (as far as I got before a bugged migrant were trashed the party: the bugger refused to move until it turned [the previous migrant were was safely contained]). Haven't tried raiding, though.
  Further, a civ that ought to have been dead could be made dead by setting unk11 and unreliable_lost_last_site, but only when no sites actually remained (the experiment where a site did remain still embarked as a live civ).

Thus, it seems you can actually revive dead civs if you're prepared to hack these flags when they start out at suitable values.

The build up of the first site was done with 0.47.04 (LNP -r04), while the flag hacking and embark was done with 0.47.05 (LNP -r02).