There isn’t much to say. The four levels are, in decreasing order or severity:
ERROR: anything that causes a task to fail is reported as an error. These may not always be critical but since they always prevent a task from completing properly, they are logged with the highest priority. That level corresponds to the Task failure Log Level.
WARNING: anything that does not cause a task to fail, yet has an impact on how the task behaves or performs is logged at that level.
INFO: all major functions performed by each task are logged with this level. This level contains most of the information you would need to determine if a task is performing what you expect from it.
WARNING, INFO and ERROR events are logged when setting the Log level to Task success and failure with details
DEBUG: contains extra messages that detail several intermediate steps while performing a task. As its name implies, this is usually for debugging purposes and is not required for normal operation. However, because some issues may only be detected long after they occurred, it is often better to have debug-level logs enabled. These are verbose, but they contain as much information as possible for each task.
DEBUG, WARNING, INFO and ERROR events are logged when setting the Log level to All events with details
One thing to note is that there is a recent feature that allows you to prevent the logging of any message in a process unless an error occurs (in which case, the full set of messages for the process - as set in your preferences - is written to the log). Double-click on a process and tick the Minimal logs on success option. The idea for this feature is that once you’ve thoroughly tested a process, or when a process has been running for months without any issue, you probably don’t need the logs to contain all the entries because they are pretty much useless (since the process is running as expected).
This feature allows you to set your overall logging level to DEBUG but to turn off that logging on as many of these well-tested processes as you see fit, knowing that if they ever run into an error, then the logging will still occur. The feature helps keep the size of the logs to a manageable level and makes them easier to inspect if an issue pops up at any time.