I don't know what they're doing, but highly-denormalized tables are very common, and sometimes even the norm, depending on how things are set up, especially in OLAP contexts.
Highly denormalized tables are often the norm simply because the tables weren’t properly normalized to begin with, and the data model wasn’t properly done, such that reasonable joins are overly difficult.
OLAP is of course its own problem, and most of the best practices for OLTP do not apply.
