Now, the census question have always been political. Ironically, the federal government doesn't survey stakeholders, in any scientific way, about what they need in the census. And frankly, the census would be many times longer, and compliance would be enforced my the army, if you asked the average business owner or city official. The city of Wood Buffalo, for example, actually does a yearly municipal census (at a cost of tens of thousands of dollars) because they feel they can't wait for the federal one (every 5 years).
The Canadian census isn't a tool to be used exclusively by the federal government, but it is a tool that investors and municipal and provincial officials used every day. Sewers, Italian restaurants, schools, radio stations, etc, are all located based on the census.
Individual Canadians benefit from the census as well, as it drastically reduces the number of surveys they are asked to respond to.
Some of the questions on the long form census are of dubious value (such as knowledge* of a second language), and practicality dictates that you can't ask too many questions, otherwise people will stop answering the most basic one (how many people live in your dwelling). So certainly the questions should be up to debate.
Making the long form voluntary will discredit the results. Not, perhaps, at the national level, but definitely in neighbourhoods. So, expect more phone and online surveys, more Italian restaurants in odd locations and more half empty schools. Cost: billions.
*People one the long form census are asked to self asses whether they are fluent in their second language, something that is highly subjective and variable by location and culture (good luck understanding the 63% of Edmundstonians who say they "know" English as a second language (only 5% of the Edmundston, New Brunswick, population has English as a first language learned and still understood). I would argue most English Canadians "know" French by the standards of knowing English in Edmundston.