That is a known bug - the workaround is to add/remove (or remove/add) the proxy role (any role, but that one is safest 
)
I think this API call will fix it too, though I haven't personally tried it: https://www.elastic.co/guide/en/cloud-enterprise/current/Platform_-_Runners.html#resync-runner
I think the trigger for the bug is a host reboot maybe? So not that common. I don't think we've fixed it yet but it's on our list...