SQLite interruption seem to put Corrosion in a bad state

[jeromegn]

There's a 30s timeout for releasing a checked in write SQLite connection. This is to prevent any one write query from blocking every other.

The way this is enforced is by interrupting the connection if it is not returned to the pool within the allotted time.

Somehow, this does not work very well and a restart is usually (not thoroughly tested) required for operations to resume properly.

Investigating why this happens and fixing it would be ideal.