But Einstein only needed one journal editor to decide that his paper was good stuff that would rock the boat, whereas under peer review, he would in practice need every peer reviewer to agree that his papers did not rock the boat.

The exact rules of peer review vary between different journals and conferences, but in general no single referee has veto power. If there is major disagreement between referees, they will discuss, and if they fail to form a consensus the journal editors / conference chairmen will step in and make the final decision, after possibly recruiting additional referees.

This seems to be a more accurate process than having a single editor making a decision based on only their own expertise.

Recall the recent study reported in nature that only three of fifty results in cancer research were replicable.

That's a false positive problem, while you seemed to be arguing that peer review generated too many false negatives.

Anyway, neither referees nor editors try to replicate experimental results while reviewing a paper. That's not the goal of the review process.

The review process is not intended to be a scientific "truth" certification. It is intended to ensure that a paper is innovative, clearly written, easy to place in the context of the research in its field, doesn't contain glaring methodological errors and is described in sufficient detail to allow experimental replication. Replication is something that is done by independent researchers after the paper is published.