As noted by @rwblickhan unknown containers running on the same VPS as inertia can cause unexpected behaviours, notably that Inertia usually cleans up everything if something fails.

I will be putting in a patch for v0.5.x but this will need to be forwardported for the significantly reworked configuration and client that has been done since v0.5.2. A forwardport should have a better implementation - maybe an array of ignored_containers

• hotfix and release v0.5.3 (PR: config: hack in an ignore-death option #608)
• forwardport to master for v0.6.x releases (related: deploy: history update can fail if container starts slowly, makes superfluous header write #626, pr: chore: bump to docker-compose/1.25.4 #649)
• come up with a longer-term solution to this maybe? (see: provision, daemon: ECS AMI's have a new ecs-agent, makes inertia confused #633, provision, daemon: ECS AMI's have a new ecs-agent, makes inertia confused #633 (comment))