One folder keeps getting reverted to a really old state

I have a single folder containing my KeePassXC database that keeps being reverted to an old state from 2022. Everything in the folder that was created since is removed (mostly sync conflict files) and the one file I care about is reverted to an old state. The only trace of this old file that I was able to find is a sync conflict from May 2022 (the oldest conflict file) which keeps getting deleted during this event but is somehow restored by other nodes).

This has happened a total of 4 times in the past few weeks. 3 of those happened right after booting my laptop (before the last event I enable trashcan versioning, so I can see that everything was moved there 2 minutes after boot) and one of them happened to my workstation, presumably because the changes were propagated from the laptop.

I have the following nodes sharing this folder:

  • A laptop running v1.27.12 on Fedora Kinoite 40, this is where the issue happens most often
  • A desktop running v1.27.12 on Gentoo
  • A desktop running v1.27.12 on Fedora KDE 40
  • An always-online server running v1.27.12 on Fedora IoT
  • An always-online server running v1.28.0 inside the linuxserver.io container on Fedora IoT ARM.

Any ideas what might be causing this file to be reverted? Should I just nuke the share on all nodes and start over?

Syncthing doesn’t keep old state around, so some device must be pushing this as the latest and greatest. You don’t really describe your setup, but I would expect this to come from some device set as send-only (so not accepting changes from other devices) where you, or someone, hits “override changes” (thus pushing a possibly old state onto everyone else).

1 Like

Every node is set to send and receive and they all have the latest version, I double checked it after the first incident. The nodes either have trashcan or no versioning. There is no sign of any conflict in the Web UI of any of the nodes.

The only instance of that old file that I managed to dig up was a sync conflict, the oldest of them present. I can see this conflict file in the trashcan and in the daily backups, but it doesn’t show up in the folder itself after an incident, which is what lead me to believe that the whole state is rolled back, not just one file.