(Speaking in personal capacity here.)
There exists a continuum of hiring practices which place more or less burden upon candidates. I would encourage HNers with input into their company hiring practices to choose less burdensome approaches over more burdensome approaches, particularly when less burdensome approaches also yield additional signal about ability.
A thing you can expect to be asked in 2016: you can expect to be asked to have an on-site interview in a city you do not live in. (How many people will be asked to do this for the SFBA or NYC? Literally jumbo jets full, this week.) For some candidates, this will involve more than 16 hours on planes, a (minimally) overnight stay in a business hotel, and six solid hours of interviews in the course of an 8~10 hour day. This requires minimally 2~3 days of these candidates' lives.
Most companies keep "conversion rate between onsites and job offers" very close to their vest. To put it mildly: it is not guaranteed that if you're flown out you will get an offer.
If you are hypothetically designing your interview process, and you replace the onsite with even an excessively long project, that's a win. The project can be written at the candidate's own pace and schedules conveniently around their other obligations. They do not have to arrange child care, take off time from work, or make tradeoffs like "Do I do the project or do I attend a friend's wedding?"
If you fly across an ocean and start bombing an interview, that's a terrible result for everyone. If you happen to be doing a project and discover "Oh, wait a minute, I've really miscalibrated here: this is far above my level of expertise with $FOO and honestly if this is the character of the work then I'm not sure I want to do it", then you have an easy option: simply close the window and, maybe, send a two-sentence email to the person who you were talking to.
Smart use of projects as a filtering mechanism can minimize costs to candidates and the company of administering high-cost testing (e.g. onsites, long projects, etc) to candidates who will ultimately not be successful at receiving offers and/or defer the high-cost assessments until the anticipated chance of receiving an offer is "very high."