It's not really about standard compliance, it's more an issue of how the browser behaves in non-standard situations and other fun edge cases. If devs only test with chrome then it's very easy to become dependent on behavior specific to chrome.

For example, yesterday I was implementing a server that uses http multipart to stream images to a browser. multipart uses two methods to frame messages: a length prefix and a boundary token. So when you send a message you prefix it with something like

    Content-Type: image/jpeg
    Content-Length: 12345

And you end it with something like

    --MagicArbitraryBoundaryTokenPreviouslySpecifiedInHead

In principle these are redundant, you could frame the content with either one but both are provided. And that's where the problem started. I misspelled a property name and because I'm using javascript it helpfully inserted 'undefined' in the "Content-Length" header. Firefox refused to display the images but chrome happily displayed them as if there were no problem.

If I had only tested in chrome I probably wouldn't have noticed that my code had a problem that was breaking other browsers.