When single batch jobs fail only the last one is party shown in the Apex Logs page. This is not enough for debugging.
A better way ist to check the log entries that are written per batch execute.
The problem is: this does not work reliably as:
- Logs seem to be written only when you have an open Developer console
- The Developer console does not fail to reload due to strange "Server timeout errors"
- The Developer console doesn't stop keep track of the new logs (sometimes it just stoppes without "Resume Updating" being available.
How can I guarantee that I have ALL the logs produced by a 10h batch available when I log in after it finished?