Recreate database failure with Canary 2.2.0.105

A bug has appeared in this version that I didn’t have before:

When I delete the database and have it recreated, it works without any error messages, but afterwards the backup jobs are no longer executed:

2026-03-01 11:56:18 +01 - [Error-Duplicati.Library.Main.Operation.BackupHandler-FatalError]: Fatal error
DatabaseInconsistencyException: Found inconsistency in the following files while validating database:
D:\Python\Python-Automatisierung\VC Code\test.py, actual size 0, dbsize 158, blocksetid: 19057
D:\Python\Python-Automatisierung\Kursmaterialien\09 - jupyter\venv\Lib\site-packages\win32comext\axscript\Demos\client\wsh\blank.pys, actual size 0, dbsize 158, blocksetid: 19057
D:\Python\Python-Automatisierung\Kursmaterialien\09 - jupyter\venv\Lib\site-packages\soupsieve-2.3.2.post1.dist-info\entry_points.txt, actual size 0, dbsize 158, blocksetid: 19057
D:\Python\Python-Automatisierung\Kursmaterialien\09 - jupyter\venv\Lib\site-packages\nbclassic\bundler\tests\resources\subdir\subsubdir\.gitkeep, actual size 0, dbsize 158, blocksetid: 19057
D:\Python\Python-Automatisierung\Kursmaterialien\09 - jupyter\venv\Lib\site-packages\jedi\third_party\typeshed\stdlib\2\distutils\command\sdist.pyi, actual size 0, dbsize 158, blocksetid: 19057
... and 29 more. Run repair to fix it.

Background:

- I perform this action from time to time to prevent corruption in the backups. At least a few years ago, I found this necessary with Duplicati.

- After experiencing this effect with some jobs that ran smoothly for a few months and where recreating the database always worked, I have now completely restarted two backups because of the error. The jobs ran two or three times. Now I have recreated the databases, and the error appears immediately. All attempts at repair were unsuccessful.

What this means is that the database has a filesize of 0 bytes, but should be 158 bytes.
You can see that all entries are pointing to blocksetId = 19057, so my guess is that this block is somehow empty, but should contain 158 bytes of data. Since the same entry is present for many files that should not be identical, this points to some file metadata that is botched somehow.

Can you create a bugreport database, upload it somewhere, and share a link with me? (You can PM me).