I stumbled on an interesting thing yesterday while playing… :))
Seems that getting all data from a table is faster if we add a "phantom" where part
that forces a clustered index seek instead of a scan.

More here.