Of course we all know what will really happen is Google will continue to use Chrome to obfuscate the way the internet works and all of the big tech companies will sell "solutions" to the problems they're causing. Most people will end up completely dependent on Google's Advanced Protection Program, Microsoft's SmartScreen Filter, etc..
Yes, the solution is giving users adequate information. That means telling them whether it's a link or a file, whether the file will execute, or what program will open it.
I think this is a "further research needed". For starters, how often will these situations actually occur in the first place? This may be less often than feared. How many users will be confused? This may be less than one might assume. Is just downloading a .zip file actually a security risk or "merely" confusing? That would make a big difference. What design will help with that? This may be counter-intuitive.
There's also some that are less commonly encountered by non-tech people, but still fairly common in the tech world (depending on the person of course): .sh, .rs, .cab, .java, .py, .org, .pl, .pm, .target, .pub
That's just from a quick check from what I recognized from memory; I didn't feel like writing the code to also print out the MIME type and description and there may be more. Here's a full list of conflicts I generated from my /usr/share/mime/globs and the TLD list from ICANN, 74 conflicts in total: https://gist.github.com/arp242/ca24b58eaf37184b03e626c8e4093...
Is .zip more common? No doubt, so it does become worse, but this confusion already exists, and will no doubt grow in the future with further proliferation of tlds and file extensions.
Basically, we should probably do something about this regardless, although I'm not entirely sure what.