Weird sync behavior, bug or operator error

Yesterday I filled out items in my daily journal template on my “desktop” Mac. Later on in the day I sent a web link to Logseq using the iOS share sheet. It gets appended in the journal for that day and tagged [[quick capture]]. Not sure if I used Logseq again after that.

This morning I discover that my journal entries from yesterday are corrupted with duplicate blocks and children. And child blocks missing or split between the duplicate and the original block.

I remember some history feature that allowed me to restore a corrupt page after sync, but maybe it wasn’t a journal page? My only option was to hopefully find a backup in my graph directories. I did something similar before, but I’m essentially figuring it out as I go. Found the directory ~/Documents/Logseq//logseq/version-files/local/journals/ with a date named group of folders. Found the folder for my corrupted day and the file that was uncorrupted.

I then quit Logseq, renamed the corrupted file in //journals directory and moved a copy of the uncorrupted journal version and renamed that to match the naming convention. (Please not that I have not yet updated all my file names. I just discovered the recommendation to do that in Logseq settings yesterday.)

Relaunched Logseq, scrolled my Journal view back from today (the 6th) to yesterday (the 5th). Yay! My correct journal outline was all there. Then, while I was watching, the corrupted version of yesterday’s journal was restored. Right before my eyes! Boo!!!

All other daily journal pages seem to be fine. I’ve shut down Logseq to write this help request and go for a walk before I break something in anger.

What am I doing wrong?
What do I not understand correctly?
Is this a bug?

I will hack the MD files to get the content I need into a new journal page and let the corrupted one remain that way. I guess I could edit the journal file that Logseq wants to mess up directly. Delete the corrupt content, and past in the correct content from my backup.

It occurs to me that maybe I needed to reindex the graph. But how do I do that when the recorruption happened almost immediately?

Thanks for your thoughts.