This case occurred in a couple of different forms. One was an appendix that was an html file that was a formatted survey form. This documentation shows how to extract the html survey.
It is a horrible mess of markup used to make a pretty looking result but, it is very difficult to make it accessible. Throw up your hands and encourage the opinion research companies to find a way to make this simple or accessible.
Another form of the questionnaire shows in this report. Scroll down to the appendix and you will get accessible html but the questions have been flattened so much that they are hard to navigate. I threw up my hands and left it as is for illustration of an area needing improvement earlier in the survey process.
One of the files in the dataset:
is a tremendously complex multiple worksheet spreadsheet. Each table is also beyond the WCAG limits of comprehedability, so CSV is the only possible format. It is the output of a very comprehensive analysis program. The referenced .csv file appears to be raw input and would have to suffice for archive.