If we can't access the robots.txt (say the server is unreachable or
returns a 5xx error result code) then we won't crawl the website at
all. So in other words, if anything, the robots.txt file should be
reachable if you want the website to be crawled and indexed :).
Usually if the robots.txt is unreachable or returns a server side
error, then the rest of the site is similarly inaccessible though.

Looking a bit further, if your server goes down or you are doing
maintenance and you need to use the robots.txt file to control
crawling of the website, then you MUST make sure not to return the
"The server is down" page with a 200 result code. Doing that could
result in us seeing this page instead of the robots.txt page, which
could result in us crawling and indexing content that we shouldn't
once the site is back up. Optimally, all such "the server is down"
URLs should return result code 503 (service unavailable). Doing that
also prevents search engines from crawling and indexing the error
pages :-). Sometimes I'm surprised at how many large sites forget to
do this...