Yeah, wal can also grow very large, making read operations increasingly slow on wrong checkpoint configuration, which also makes checkpointing slower and it grows out of proportion.
Inefficient queries, no reader gaps and a bad manual checkpointing system can bring the system down.
That's neat, I haven't worked with SQLLite for a few years but this changes it.
Of course you can still fall into the trap of reaching your checkpoints faster than the first wal file needs to write back to the database, but I think that's a less critical problem in pratice.
Inefficient queries, no reader gaps and a bad manual checkpointing system can bring the system down.