What specifically do you mean by "book review?" The term gets used for student projects/tasks as well as professional historians' work, so it's helpful to differentiate; apologies if I'm addressing it in a way not relevant to you.

I can speak more to the latter than the former, because to me the former felt comparatively like an exercise in trying to write a book report that included a demonstration of things I knew from a seminar than an actual review in the field. (And yes, it took a lot longer than they do now.)

For writing reviews for journals, the time factor depends on three things: the journal in question, the length of the review contracted, and the proximity of the book to my own narrow specialization. If the journal is more prominent, the length more significant, the specialization closer, or any combination of the three, it will take longer than without, because I will have to go into details and historiographical debates, matters of detail, and concerns about the audience and so forth.

The important things for me to figure out are as follows (assuming a normative female author):

• What does the author claim she is trying to do?
• What does she actually achieve?
• Is there a major set of debates she intervenes in, and how? (Here the distinction between what she presents in her introduction and what she actually does by implication is important.)
• Is her source material good? Is her analysis good? Is anything big missing--major interventions or source matter that would complicate her work? (A few examples, or egregious points in the negative case, are called for here.)
• Is the book presented well in terms of structure and language? How is it structured?
• Who is the book written for, and does it connect on the proper level? Who should make sure they own it, or read it?
• As a subordinate point, does the editor or the press just suck or totally rock in terms of production values, indexing, citations, and other mechanical things? How useful is this book?

There are others you could put in here. For example, I recently reviewed a book on late apartheid South Africa (a bit afield from me), of medium length, for the major journal in my field. It took me about eight days of off-again, on-again work--maybe 20 hours all told? Some academics pop out reviews in an evening, by simply identifying the thesis, summarizing, and noting any really huge gaffes or problems, but I don't think that does anyone a service. Another review I wrote, about 1500 words on a book right smack in my subfield for a technical journal, I spent at least three times as long working on. Just keep in mind what the readership is going to want to know about the book, and be honest but fair (because a book is a hell of a lot of work to put out)--you need not pick at every single thread, just the ones that are likely to flag the interest of the readers who will look at your review as a guide to whether the book will be worth consulting.

I hope this is helpful somehow. Honestly, a lot of it came with practice and reading enough other reviews to see what's included or excluded. A few of my colleagues actually crib notes from other reviews of the book as well to see if they concur or differ with the points other reviewers raised, which strikes me as unsporting if that information goes into the review uncited.

[edit: I mess up bulleted lists in reddit grammar. WOE IS ME. I also added the last paragraph there, because I'm a chatty, chatty person.]