Why a document-by-document read fails
The most common problems are contradictions between documents that each look correct on their own: the same control stated with two different figures, a lifting operation or item of plant with no corresponding method statement, a substance named with no COSHH assessment in the pack, a method statement citing a revision of a risk assessment that is not the one supplied, and competence records that have expired before the works start.
None of those are visible while reading a single file. They only appear when the pack is treated as one submission and the documents are compared against each other. That is the whole basis of the checklist below.
Before you start
Two things make the review worth doing. The pack should be complete, because reviewing half a submission mostly produces a list of things you already knew were missing. And the reviewer should be the person who can accept or reject it, or be reporting directly to them, because a review nobody acts on is an expensive way to produce a document.