Failed to acquire the synchronization lock. etcdserver: leader changed

[static-moonlight]

Sometimes (not that often, but it happens), I get this error:

Failed to acquire the synchronization lock. etcdserver: leader changed

1. I'm not really sure what happened to the etcd leader, and why it changed, or why (occasionally) only just 1 workflow is affected by that, and not any of the the others. Can someone elaborate on this?
2. The workflow state is weird. In the workflow overview it is marked as failed:
   
   ... however, when I try to restart it manually, i get an error: Internal Server Error: Workflow cannot be retried with node [...] in Running phase. Sooo, what's the actual state of this thing? Is it dead or alive?
3. I'd like to automate as much as possible. Meaning: if something like this happens, I want Argo to just retry the workflow, without manual interaction. Is that possible? If yes, how?

I'm using Argo Workflows v3.5.11.

Note: The workflow will be around for another 4 days, before it's automatically removed. So I can provide additional information until then.

[static-moonlight]

I'm still having this problem occasionally. Does anyone know something about this?