Took some time to go over your comment. All of your points are valid and what they truly show is how complicated a decentralised system is. All of the different apps using their own slight variations makes for these annoying exceptions. 

As for "render the page on Steemit and inspect that for valid image tags with something like BeautifulSoup" that sounds interesting. I have no experience with this. 

How I currently deal with the issue is basically verifying each post beforehand and filtering out any 1% by hand ;)