I agree completely that testing task knowledge is superior to testing syntax, and I'd suggest that testing syntax is more tightly tied to selling "stuff" than testing task knowledge. After all, task knowledge doesn't need you to recertify for every new version of the product... syntax (and its GUI variants) can!

Taken to the logical conclusion, this brings us to a very interesting place, however, as many of the differences between DB2, MariaDB/MySQL, Oracle, PostgreSQL, SQL Server are syntax differences, and very mild ones compared to the differences between C++, Lisp, and Perl (for example).

SELECT Col1 FROM Table1 WHERE Col2 = 33

for instance, is near-identical on any major relational database based (more strictly or less strictly) on ANSI SQL standards. ACID compliance is the same. Normal forms are the same. The idea of transaction log/redo log files/archive logs/etc. is very similar, with twists on the implementations. The idea of Point in Time Restores is the same.

The idea of data order vs. separate indexes is identical. Compression tradeoffs are similar, and so on.

Thus, we could start with a SQL DBA cert not related to any particular product, and then move into the various flavors - and in the doing, all of us would be exposed to different ways of implementing those concepts, and perhaps even do better in the future.