This is an almost inextricably linked effect of allowing anyone who thinks (or wishes) they have the skills to compete for a $250K/yr job.
If you put a grand prize out there, lots of people are motivated to go after it, including the ultra-long shots.
On of my favorite interviews was for Acclaim in the mid-90s. Take home (download a zip file with a skeleton app, read the code, write the strategy element that’s stubbed out). I don’t recall how much time they allowed but it was at least 24 hours. Mine was “navigate an airplane from left of screen to right of screen against a fixed set of blocked squares[mountains]”
Probably spent a little over an hour on it, enjoyed doing it, and remember it 25 years later. They can score such a test (at least initially) totally mechanically, weeding out the “couldn’t get anything to work” subset of the population in zero minutes of engineer time per new application. It tests something quite close to what you’d do as a game programmer for them. You can do it at home, at your own pace, in your normal environment. If you pass, then you might have to get an airplane and travel to interview, which is a multi-day disruption to your life.
I’m on the employer side now; I see the value in asymmetric tests at the initial filtering stage for entry-level, because I don’t see any workable alternatives that don’t introduce other, likely worse, forms of credential-checking.
I’m sure we miss some good candidates this way; false negatives are bad but less costly than false positives.