I got this idea from OpenBSD Daily. Combine this with another idea I read elsewhere about reading the "upstream" dependencies of your projects, we have a winning combintation for getting relly good at the stuff we ought to know.
How well do we really know the datastores we use? the ORM library? Kafka? Reading something like the query planner of PostgreSQL itself could be an enlightening experience.
How about something more fundamental -- like reading the Collection libraries of Scala (or your favourite language?).
This is a great exercise.