I was about to blame the problem on the fact that using the YQL in this way pulls data from user-generated tables and is not Yahoos own self-generated data, which it will tell you. So they try to absolve themselves of responsibility from this.
However, just looking at the the output from their console, this returns well-formed XML. Looking at the actual supposed REST query for the same operation, it's quite obvious that the return of the query is not well-formed due to lack of definition the "yahoo" namespace.
This output is also going to give pretty much any XML parser a hissy-fit and/or unpredictable results because of the complete lack of version details, document type and encoding information that any normal XML response would generally contain.
So yes, you are correct in stating that there is probably some nutty kind of anomaly going on with Yahoo's YQL service.
Based on an observation you can make for yourself about the state of the YQL forums themselves, which seem to lack any sort of robust activity, and another looking their blog, which hasn't posted anything in almost 2 years, one perhaps might not be far from the actual truth in stating that it's most likely Yahoo is silently deprecating/discouraging use of this API and trying to phase it out.
If you plan on pulling data via YQL REST queries, you're going to need to be able to be aware of these differences, and work accordingly around them, most likely by pulling the entire query into a string rather than trying to use generic XML parsers, using something like regular expressions to insert and define
- the appropriate proper document type and encoding information and
- the "yahoo" xml name space
If you do that, then give that to an XML parser you should be fine. But try and be aware of these "difficulties" when working with the YQL API.