I doubt it would harm you, but it is a bit unorthodox. Why not make the table header a header tag?

I'd argue with the "take up space unnecessarily" point a little. Headers aren't just for SEO (their SEO impact is probably pretty small these days) - they're for visitors. If these pages get direct visits (including from search), a prominent header can really help people know their on the right path. Breadcrumbs are great for people who are already on your site and have a sense of it, but they're too small and complex for that 5-second test of "Am I in the right place?"

Generally agree, although I'd just add that Robots.txt also isn't so great at removing content that's already been indexed (it's better at prevention). So, I find that it's not just not ideal - it sometimes doesn't even work in these cases.

Rel-canonical is generally a good bet, and it should go on the duplicate (you can actually put it on both, although it's not necessary).

Just a follow-up to my endorsement. It looks like Screaming Frog will let you control the number of pages crawled per second, but to do a full crawl you'll need to get the paid version (the free version only crawls 500 URLs):

http://www.screamingfrog.co.uk/seo-spider/

It's a good tool, and nice to have around, IMO.

If you're already use rel-canonical, then there's really no reason to also block the parameter. Rel-canonical will preserve any link-juice, and will also keep the page available to visitors (unlike a 301-redirect).

Are you seeing a lot of these pages indexed (i.e. is the canonical tag not working)? You could block the parameter in that case, but my gut reaction is that it's unnecessary and probably counter-productive. Google may just need time to de-index (it can be a slow process).

I suspect that Google passes some link-juice through blocked parameters and treats it more like a canonical, but it may be situational and I haven't seen good data on that. So many things in Google Webmaster Tools end up being a bit of a black box. Typically, I view it as a last resort.