Problems with indexing start and performance in different versions - stuck celery task

[hoooli]

I installed Onyx version 0.18, but the search in the table didn't work properly (neither through the Confluence connector nor the file connector where I used the .md format). (any tips for this?)

So I deleted the containers and volumes and installed version 0.19, hoping it would be better.

Unfortunately, after this operation, the connectors in both versions stopped indexing automatically after creation. It looks like something is stuck. When I add a new connector, I have to manually trigger full reindexing — it just doesn't start.
If I try to change the embedding model, the process doesn't start at all and remains stuck.

Do I need to delete some hidden volumes outside of Docker volumes? It looks like there's a problem with Celery, as every connector eventually reports an error that it can't index due to a missing Celery task.
Even though I use different images and volumes, the situation repeats in all versions now — the connectors don't start indexing, and everything is slow, making Onyx completely unusable.
Any idea how to fix this when using force-recreate and deleting all volumes didn't solve the problem? I even restarted the entire Docker without luck.
There is no error in the log after creating the connector, and also no error after changing the model — just info about the change, and then nothing new about the task failing. I also tried FLUSHALL in Redis, but it didn't help.

Thanks