It would be interesting to see whether a team meeting held in minecraft negated the issues with video latency. If there is no expectation that you can read people's facial expressions then the latency becomes a non-issue.

I think that could help some, since the latency problem is likely made worse by the difficulty reading people's body language to identify the flow of the conversation (e.g. latency + not being able to read the body language of when someone is finished speaking means people interrupt each other).

However, I don't think it would help with the latency itself.

How would that help? VR is not a networking technology.