While it doesn’t address the topic of consciousness, you might find some of how this sort of “backwards in time change” is being applied today interesting:
Which led to what’s currently my favorite titled paper, Stable facts, relative facts: arxiv.org/abs/2006.15543
So one of the challenges that would arise from layers of delayed/hidden observations would be whether you’d even have universal agreement at the final review. i.e. The computer might have observed the cat as alive and baked a cake celebrating it, but then you open the box to a dead cat, each having correctly observed a result, just separated enough that they didn’t need to agree.