Below I reproduce a modified version of a comment I left at BCL regarding a costs as a hidden rationale for eliminating the long form. I think it's worth some discussion.
I don't think the individually mailed out long form would incur a great deal of extra cost outside of extra printing and postage on the front end for much of the nation. However, there are places where the long form is used almost exclusively and administered in person by enumerators for sampling validity in terms of both response rate and population characteristics. These include remoteness, community size, existence of informal dwellings (such as occupied cabins), cultural considerations (eg. First Nations), and general capacity of individuals to respond (eg. relative literacy rates). Much of the North falls under these descriptions. Conducting the census in these places is extremely costly because staff must be recruited, trained, flown, housed, and fed for as long as is needed to visit every dwelling physically possible. For a government looking for an easy cost to cut, it would make a tempting target.
Another thought is that processing all that data also incurs substantial costs. Statistics Canada implemented a new system of data collection and processing during the last Census. That was what the hullaballoo over Lockheed Martin was all about. I have reason to suspect the performance of that system may have been a bit less than advertised, and not worth the cost. Replacing it might be even more costly especially at this late stage (next Census is next year), so why not cut out the expensive bits yet again? Again a juicy fiscal target.
Whatever the case, I do not believe we're being given the full story.