The short answer is that your language needs === when it fucked up the semantics of ==, but it's also too popular and you can't fix it without breaking half the web.
One neat feature is you can compare to both null and undefined at the same time, without other falsey values giving false positives. Although that's not necessary as often now that we have nullish coalescing and optional chaining.
I just tested and Terser will convert v === null || v === undefined to null==v. Personally I would prefer to read the code that explicitly shows that it is checking for both and let my minifier/optimizer worry about generating compact code.
It's also important if you're checking hashes (at least, it was - if you're using correct hashing algorithm that isn't ancient, you will not have this problem).
Because if you take for example "0e462097431906509019562988736854" (which is md5("240610708"), but also applicable to most other hashing algorithms that hash to a hex string), if("0e462097431906509019562988736854" == 0) is true. So any other data that hashes to any variantion of "0e[1-9]+" will pass the check, for example:
I did use md5 as an example because the strings are pretty short, but it's applicable to a whole lot of other hashes. And the problem is that if you use one of the strings that hash to a magic hash in a vulnerable site, it will pass the password check for any user who's password also hashes to a magic hash. There's not really a high chance of that happening, but there's still a lot of hashes that do hash to it.
Like == but more strict. The == operator will do type conversion, so 0 == '' will actually be true, as an example. Sometimes (honestly, most times) you may want to compare more strictly.