As I understand it, loop guard will put a non-designated port into a loop-inconsistent state if no BPDU is received after the max-age timeout, hence preventing that port from transitioning into a forwarding state. What happens if there is a genuine loss of BPDU - the upstream switch is indeed dead, would this stop STP from re-formation?
I'm not exactly sure of what you mean by a "genuine" loss of BPDU -- genuine as opposed to...what?
Whether an alternate or backup, non-designated port, that is in the blocking state, stops receiving BPDUs because a link failed or because the switch at the other end has blown up, Loop Guard will prevent that port from transitioning to forwarding until it starts once again receiving BPDUs. This is inconsequential to the flow of traffic along the converged path because the port was blocking already anyway. It does, however, prevent the re-convergence of STP and the subsequent corruption of that primary traffic path.
If the port that stops receiving the BPDUs is the root, non-designated port, that is in the forwarding state, Loop guard will, obviously, not effect that port because it was already forwarding. It will also not effect the alternate or backup port in the blocking state from transitioning into the forwarding state once the inferior BPDUs it is receiving become the superior BPDUs, as a result of the loss of the superior BPDUs.