If it’s taking over 2 hours to setup the environment, then either their stack is so complex and poorly documented that it should be a red flag to bail out, or you grossly unqualified. Either way, doesn’t seem like a good fit for either side.
I would agree that companies that give "2-3 hour" take home problems are a red flag for me — red as in, I won't do it anymore. That situation becomes an arms race quickly: if I can solve it in 3 hours, why not spend another 3 hours doing a second, more elegant solution that makes me look even smarter? I'm sure that's what the other candidates I'm competing with are doing. And if they're spending 6 hours, I should spend 8 hours at least, to get an edge over them. But they know I'm spending 8 hours, so they're probably spending 10...
I wouldn't say simply having at 2-3 hour take home by itself is red flag. But if the take home is a pile of hot garbage (poorly or inadequate documentation on how to setup a local env for example) then that is red flag.
My first hire (that wasn't someone I knew/previously worked with) I did not require that they complete a take home test. That hire turned out to be a complete disaster. I have since started requiring take home tests, and I will never again hire anyone that I don't know without a take home test. After requiring take home tests the candidates that have been hired have been much better.
Granted, sometimes fluency in the XYZ of their stack matters - but usually it just doesn't, and grilling people for this fluency shows a lack of sense for the broader picture, in my view.
https://versastudio.com/blog/if-carpenters-were-hired-like-p...
It doesn't have to be a docker run.
>I can think of several well documented setups that can take upward of 2 hours
If ALL you have is "well documented" setups that take 2 hours, then that company sucks.
Look you don't have to have docker, but if you have a setup that is merely a Confluence page, or a pdf file, and don't have ANY shell scripts, or any way whatsoever to bootstrap setting up your environment locally, then that is a massive red flag.
"It works on my machine!"