Currently, no block elements (lists, tables, code blocks, interpreter blocks, etc) work inside table cells. They interpret their content as inline only. So this is the expected behavior.

I believe there is no product reason we can't allow tables to contain block elements, but that supporting them is complicated.

The parser currently assumes that it can find the boundaries of a block without parsing the block. When the parser sees a <table>, it can look forward to find the next </table>, and that is always the end of the block.

This isn't exactly true, and we get cases like this arguably-wrong right now, but it's broadly close to expected behavior:

<table><tr><td>`</table>`</td></tr></table>

This will no longer be true if a table may contain other block elements (like code blocks, other tables, interpreter blocks, and literal blocks) because all of these may contain the substring </table> without closing the table. Supporting this likely requires substantial work on the parser because the way it processes blocks must change completely.

This may not be enormously complex, but I'm not aware of any situations where putting block elements inside other block elements is hugely valuable, so it is difficult to prioritize.