No, because it enables the user-agent to paginate content before the entire page has even loaded.
And presumably the page author knows a priori what the header and footer for the table are going to be, so why not transmit that information to the user-agent as soon as possible?
If you examine XSL-FO, they have similar constructs for what I am guessing are similar reasons.
