I think one of the problems is that the anti-video crowd believe that Canon develop their specs first, then develop the tech & features to go in it. It's more likely the reverse. Sure, they have a rough idea of what the next camera will be, but not the details -- the market will determine that. Let's consider my guess about the time timelime of a new hypothetical 70D.
Ongoing schedule: Canon's various R&D departments continue to develop tech that could go into their crop line: several types of sensors, screens, processing chips, etc. This process never stops, they're just always trying to develop better and/or cheaper components. These same departments work on the tech for all of their models, not just the 70D, so the new screen or sensor could go into a 7DMII as well, depending on later market conditions.
About 6-9 months before product announcement, Canon must start to "lock down" their feature set. They will know from experience how long it takes to prepare a factory line for each new component and the finish assembly. Features will be locked down in the order required to lock down other components, and by the prep schedule of the factory lines. E.g., they must lock down the body size and shape before locking down layout of internal components. At this point, nearly all the tech is already designed, prototyped, tested, and estimates are made of production costs. Next, Canon checks the market conditions, and also examines tea leaves to determine what the main competition is likely to produce. From this, they select the components to go in the new product. In some cases they may be able to delay this decision for a while, but at some point they must just go with a choice or delay production.
Probably around 3 months from launch, all the features must be locked down. Now the firmware engineers start working overtime to polish and test the firmware with the exact feature set. Occasionally, the firmware engineers will find problems that can only be solved by a hardware change (e.g. processor is just not fast enough for a given task); this delays production while R&D determines the cheapest of their dev components that will do the job.
From 3 months to launch: factory lines ramp up, marketing department goes to work, writers go to work.
Based on my assumptions (yes, these are my _assumptions_) there's nowhere in the process that video can *substantially*increase cost, it's just one part of the process. Sure, it might take $10 extra or even $20 -- but nothing worth worrying about. To put it another way, they sell $150 p&s cameras that take HD video roughly on par with the SLR line (the main difference is the sensor quality), and I refuse to believe that much more than $20 is the cost of including video.
In short, considering R&D spread over multiple models and the relatively decent video quality in cheap P&S cameras, it's seems like a lot to swallow the notion that it's a significant cost in DSLR's.
Are there any production engineers out there who can critique the above hypothetical timeline?