Up until now, database instances that you opened during your work would stick
around in the engine’s memory, leading to potentially unbounded memory usage.

To fix this, we’re now cleaning up stale databases automatically: If you haven’t
accessed a database for more than one hour, we’ll close it for you. Don’t worry
though; if you access a closed database, we’ll just bring it into memory again.

You shouldn’t notice any of this — in particular, this does not delete any data!
— it just helps to keep the engine’s memory usage in check. For example, this
comes in very handy when you want to run a repeated workload by creating a new
database for every run.

Now you can issue multiple tasks, creating new databases, without stressing the
memory and your engine.