Speed (even when set at slow) is not always the problem but a regular timing of queries which is unlike humans who operate randomly.

Would it be the case of product variation (color, size, date, price)?

If that was the case you would implement rel="canonical" for the main page.

For example:

site.com/page.php remains main URL while site.com/page.php?special=january gets treated as if it was the first one.

Any links you earn through that January page would count towards the core product.

You are likely using tools like SEO Elite etc which send fast queries. This type of software can be set to slower hit rate. I doubt you guys search that fast to be booted without use of software. Check your plugins and software - that would be my first recommendation.

I do get this from time to time when I try to hack URLs to death... but it only happens occasionally 

They will be treated as seperate sites, to us www1 and www may seem very similar but to Google it will be two separate things. Even http:// and http://www. can be two different sites. So from that point of view I say you are safe given that you're not duplicating content in ways that I may not be aware currently.

Sorry this sounds a bit silly to me. I vote against.

Google understand when it's looking at an eCommerce site and it does not expect lengthy articles on each product. So if I was you I would focus on creating product descriptions as useful and unique as possible and allow customer feedback and reviews which will grow your natural content.

Focus your links towards the main domain / categories / product pages. If you do have articles use them to point relevant anchor text links (in moderation) to products they describe.

I support this view. Let's not do the same damage to our branding as Yellow Pages did to various AAA Aarwark Plumbing type businesses.

Google claims that nofollow links do not pass link juice - no matter what. If your tests show otherwise then Google is a liar (or they just don't know how they work internally). To my knowledge proper blog commenting work brings secondary value to a direct link from the page you're commenting on (scraped content links which exclude nofollow, citations, mentions, branding, networking...etc).

Honestly, no idea. What I would be "comfortable" with personally would be no more than 20.

If your network is not a secret then ensure that you have full and clear disclosure on the site(s). I remember Matt Cutts once said this as an answer to a similar question on YouTube. Private registrations are common for non-SEO purposes so Google cannot act on it alone. However if you have 1000 domains with private whois that may start looking a bit suspicious.

Hosting images externally never had any impact on cases I had a chance to observe. The only problem I can think of is that you lose control over loading times or if somebody takes an image and links (credits) the image hosting domain instead of your domain.