The above url and the test case below trigger an incremental reflow causing the problem reported. There is a separate bug (10716) addressing that, and if it gets fixed, we'll need a different way to invoke this bug. The table cell's area frame containing the radio button is returning a 0 max element width (the radio button is returning a non zero max element width). <TABLE WIDTH="100%" BORDER="0" CELLPADDING="15" CELLSPACING="0"> <TR> <TD WIDTH="100%" VALIGN="top">foo</TD> <TD BGCOLOR="#FFFFFF" WIDTH="175" VALIGN="TOP"> <table width="100%" cellspacing="0" cellpadding="0" border="0"> <tr> <td bgcolor="#669900"> <form> <input type="radio" name="vote" value="1"> </form> </td> </tr> </table> </TD> </TR> </TABLE> foobar

Marking invalid after Troy explained that max element widths are only being set when the avail width is unconstrained.

And the reason is "initial".

I don't think we currently restrict it just to 'initial' and I think 'resize' is okay, too. You raise an interesting point, and it needs clarification exactly when it applies

Based on Troy's comments, marking as verified invalid.