Type checking slows things down when extracting random data and trying to convert to "useful" types. Interacting with the PostgreSQL database is much faster when all data is extracted at strings and converted as needed, rather than by letting the pgdb.py module do it for you.
In cases like this, the programmer knows more about the type of data that is expected than the helper module, so things work faster if you can guide the type conversions.
Using open source code is obviously a huge win in these situations. Being able to poke around inside the code, profile and understand is very beneficial.