I wonder if this was a concession to the limited browser code and slow networks at the time. With the img tag you know its going to be a jpg, gif, or png. If it was a media tag then you have no idea what it is. Imagine downloading a 10mb quicktime file or flash object on dial-up so the mime header can be read as you wait for the page to render. Its only recently that anyone has bothered to handle video in the browser.

Could web servers back then just send the client the header of a file? I think these guys were working with a lot of ugly limitations that we've only recently overcome. Andreeson recently said that he expected the back and forward buttons to be temporary but no one thought of a good replacement for them. We still haven't.