I'm doing some ad hoc reviewing this round. I do try & read the instructions, and more, follow them. Here are some of the points we've been given:
Everything to be considered in the review process should be included in the application and the folder for the application materials. Do not visit any web sites cited in the application [my red emphasis] as that might jeopardize reviewer anonymity.
Aside from the implications about the saavy of people in tracking who visits their websites, this is an important message. You can (and are supposed to) link to papers in the pubmed database from your biosketch. Papers are OK. Websites are not. What you need to say needs to be in the proposal. IN THE PROPOSAL. Remember, page limits are your friends. Page limits tell you how much information needs to be in the proposal. You are not a special snowflake, doing special snowflake biology that requires 10pt font with 5mm margins because you have that much that is important to say. I surely (as a reviewer) am not interested in reading it.
you should score the applications AS THEY WERE SUBMITTED [caps from letter to me, not me], not as they would appear after even minor revisions.
Reviewers are not your friends. They are not supposed to guess "what you meant" or "of course that's something this PI would fix". Reviewers read the proposal, not your mind.
More on why reviewers are not your friends:
State only strengths and weaknesses, do not give advice about how the applications should be revised. That would not only be unnecessary it is actually inappropriate.
And in fact:
Your primary audience is the Program Officers and leadership at the funding institutes, not the applicants,
When you write a grant, reviewers are just that: reviewers. I (as is true of many applicants) tend to think of them as evil, maniacal, spawn of BSDs, unless of course they fund me, in which case they are a combination of angle, genius, and whatever-else-I'm-loving-at-the-moment. But they're not either of these things. They are poor slobs, like me, trying to do a job, which is review a very large stack of applications. One of your jobs is to make their job as easy as possible. As easy as possible.
When you get your pink sheets, remember: They are not writing to you, they are writing to the people who make the decisions, NIH staff. They are justifying their score to people who will balance their score with a thousand other factors in deciding who gets funding. This is a subtle difference from telling you how to improve your grant.