Issue with concourse worker volumes

Hello,
We have the concourse 5.8.0 deployed:

  • web node on VPS
  • worker node deployed on Windows 10 via Docker for Desktop, via docker-compose script. (4 cores, 6GB RAM, 250GB SSD). All works using the linux env inside.

We are currently using btrfs baggageclaim driver, but have tried with the overlay as well.

Our builds freezes the worker sometimes. The more work happens on the worker (and I mean more that 1 simultaneously), the higher the chance of freezing the worker.

Looking into the worker node, I see the message (which repeats every few seconds):
{"timestamp":"2020-03-24T09:07:00.017012917Z","level":" error ","source":"baggageclaim","message":"baggageclaim.api.volume-server.list-volumes.list-volumes.failed-hydrating-volume","data":{" error ":"EOF","session":"5.1.38585.1"}}

The disk is no near being full (usage is at about 40GB of 350GBs).

Aby idea what is going on, and how to improve the stability?

UPDATE: It seems that launching only two pipelines at the same time causes massive slowdowns, commands effectively freezes, and resources turns orange.