Probably they just need to adjust their algo to remove # fragments before trying to identify if the link is an image or not.
But the real way to solve this is to ping the server to download the opengraph preview and checking the content type before downloading the image.
#m=image%2Fjpeg&dim=1049x1920&blurhash=%5D55%3A%3A7%3DfR%24i_WA%7DI%24koIniaeXgt3r%40jFnj6IX3w%7DjYsCKHf%24sDWBn%2BR*a%23offkWVWVR%2BS2bIW%3AI.NaShX8W%3BNHNaS4S%23X8&x=0c1b2454008bd3fcc94dbfbe84044833dce51e7acaf5e84db4921bba01884d17