It could be my understanding of what was going on was incorrect, but I'll tell you what I observed, and please let me know if there was a better way to fix it. I found the easiest way to do so was just to avoid using Views. And frankly, wasn't surprised by this, as using UDFs w/ SQL Server also had really bad perf, so I didn't question that a particular feature was badly performing.

Ok, lets say you have two large tables: Sentences and Annotations, and you make a view that joins those, lets call it vwSentencesDetailed. Then you do a query like:

select * from vwSentencesDetailed where srcID = _srcID

Where srcID is an indexed field in the Sentences table that will let you get a narrow subset of the data. Let's say the particular file it came from....

The performance characteristic I observed was that the it behaved like (loosely):

select * from (Select * from Sentences s Inner Join Annotations a on a.sentID = s.sentID) vw where srcID = _srcID

rather than like:

select * from (Select * from Sentences s Inner Join Annotations a on a.sentID = s.sentID where srcID = _srcID) vw

I believe that SQL Server makes such a performance optimization, but Postgres does not.