I merged an improvement yesterday to avoid these span HTML elements when they come from Word, but I’ll take another look with your example. Please let us know if you find any other example.
Just note than a span element with the lang attribute is a valid combination as an inline html markdown-it token, the system is keeping them for consistency because it’s “valid”, my concern is “fixing it too much” and stripping instances with additional attributes that were intentional (maybe not even an issue, I’ll investigate).
This is not the case because the clipboard contains both HTML and text/plain in this scenario, and the system uses the appropriate origin if you paste outside vs inside a code block.