Hi @Checconio
this may have nothing to do with it but I also had a lot of memory « leak » issues before the « lockup » of the RAM allocated to DuckDB and it was still running at the time of the Gladys Plus backup (and on top of that it was also the backup of my LXC 
).
Afterwards the RAM stabilized at 2.5-3GB (like what I’ve almost always seen).
And then one day (22/12/25 to be exact), it happened again with memory leaks over several days (with Gladys repeatedly restarting) without understanding why:
And the culprit wasn’t Gladys but Z2M, I explained it here.
To put it simply, one of my modules was sending 4 payloads/second and completely saturating Gladys for me.
Since I fixed the issue by « blocking » the info sends from this module in z2m for several seconds, my Gladys has been running like a charm!
So maybe also look at the integrations and not just at the Gladys logs, which didn’t tell me anything in my case.