Everybody does not in fact know that "actually working with someone" is the "only" way to get an accurate evaluation of an engineer. There are recruiting/qualification disciplines besides temp-to-hire and whiteboard interviews.
Temp-to-hire is problematic for its own distinctive reasons:
* It's difficult to generate apples-to-apples comparisons of candidates in a temp-to-hire setting because the engineering workload isn't constant; candidates are getting the luck of the draw with whatever the tasks are during their contracting period.
* There's as much subjectivity built into the projects candidates get during their temp-to-hire as there are in whiteboard interviews, so culturally preferred candidates are just as likely to get fast-tracked in a temp-to-hire process as with interviews.
* Interview processes routinely evaluate many candidates for a single opening --- that's expected in basically all hiring, across every industry --- and can't reasonably offer temp roles many candidates simultaneously, so in effect temp-to-hire just pushes the qualification problem to a higher level of the funnel, where the recruiting process decides who to extend temp offers to.
* Temporary contracts draw out the hiring process and potentially make it take far longer to finally fill a role than other processes do; you're required to trade accuracy of evaluation against the time taken to fill a role, bearing in mind that when candidates wash out of their temp contract, you likely have to start all over again with another candidate, making it even more important to pre-screen candidates before offering roles, which just recapitulates the whole problem.
* The tech industry is notoriously bad at evaluating on-the-job performance too; every developer I've talked to has stories of dead weight team members† that were kept on the team interminably because no process existed to evaluate their performance accurately. There's a whole cottage industry of people who have failed up† through role after role, because there's no resume difference between people who phoned in a role† for a year or two and then decided to move on, and people who did well.
* Perhaps most importantly, temp-to-hire is deeply unfair to candidates; a career norm in tech is that people secure their next role while winding up their previous one (in fact, that's a norm in most white collar industries, which is why there's so much written career advice about how to handle counteroffers). Taking a temp-to-hire position requires a candidate to forego income security for a chance at a job; when that job doesn't work out, they're now in a distressed position when looking for their next job.
(I'd add, though it's a nit, that temp-to-hire positions virtually never compensate appropriately; contractor wages are routinely 2x+ what FTE wages are, in large part precisely because of what makes temp-to-hire so attractive: freedom to fire on a moment's notice with no commercial/reputational cost.)
† Shut up, Kurt