yes, there are too many variables between the candidate, job, company, and work environment to determine long-term fit via a test, especially for "creative" jobs. the more regimented the job (e.g., fast food cook), the lower the variability, but it's still significant. plus, such tests only evaluate technical skills, not the more important non-technical ones (like punctuality, integrity, steadfastness, etc.--note that these are a function of the involved parties and the relationship between them, not just the candidate).
but also, the underlying problem of hiring is not one of trying to get the best fit, but of trying to avoid the pain of firing. that's the thing that needs to be reframed/solved, but that's a much harder and a much less technical problem (alternatively put, technical tests are marginal at best).
I spend weeks drumming up two or three days worth of coding work. The coding aspect is basically manual labor and pedantic arguments with other devs.
a 10-day contract is better, since it's real work, for pay, but the relatively short duration doesn't tell you much about the candidate's intrinsic motivation or how relationships develop past the honeymoon stage.
so really, it'd be best and easiest if we all explicitly assumed that jobs had 6-12 month trial periods, for both parties, and that after that time, either can walk away without hard feelings (or negative judgment), other than in the most egregious cases (i've seen a couple cases that'd fall in this category). again, this is primarily about jobs that have the most variability. less variable jobs don't need as long of an evaluation period (but do need more than a few weeks).
my (now failed) startup in this space attempted to answer it for less variable jobs (hourly work), where we could tease out more signal from the noise, but even that had lots of unaccountable variability.