Ansible: clear host errors after handler call - ansible

Question

What I Do
For Bare Metal deployment, I configure interfaces on CentOS 7 servers via ansible 2.7.9.
Sometimes, the interface definitions change
- name: Copy sysctl and ifcfg- files from templates.
template: src={{ item.src }} dest={{ item.dest }}
with_items:
[...]
- { src: 'network.j2', dest: '/etc/sysconfig/network' }
notify:
- Restart network service
- Wait for reconnect
- Reset host errors
which is why I call a handler to restart the network.service when a change happens:
- name: Restart network service
service:
name: network
state: restarted
- name: Reset host errors
meta: clear_host_errors
- name: Wait for reconnect
wait_for_connection:
connect_timeout: 20
sleep: 5
delay: 5
timeout: 600
What I want
I can't get ansible to not quit the run when the Restart network service handler fails. Since the service restart is working fine on the host itself, I either want the restart to always exit with RC=0 or clear the host error after the failing handler call. In the following list, is there anything I am missing or doing wrong?
What I tried
ignore errors: true, failed_when: false, changed_when: false directives. Either with shell/command module or service module in the restart network handler block.
meta: clear_host_errors directly below the - name: Copy sysctl and ifcfg- files from templates. block
calling meta: clear_host_errors as a handler
Having the handler Restart network service exit with || true
async/poll variants for Restart network service
setting Pipelining to false
I always end up with:
RUNNING HANDLER [os : Restart network service] *******************************************************************
fatal: [host-redacted]: UNREACHABLE! => {"changed": false, "msg": "Failed to connect to the host via ssh: Shared connection to aa.bb.cc.dd closed.", "unreachable": true}
RUNNING HANDLER [os : Reset host errors] *************************************************************************
fatal: [host-redacted]: FAILED! => {"changed": false, "module_stderr": "Shared connection to aa.bb.cc.dd closed.\r\n", "module_stdout": "", "msg": "MODULE FAILURE\nSee stdout/stderr for the exact error", "rc": 0}
RUNNING HANDLER [os : Wait for reconnect] ************************************************************************
Where is the correct placement for meta: clear_host_errors in this case?
Some additional info
restart of network.service takes about ~40 seconds. (Tried async: 120 and poll: 30)
established, non-ansible SSH connections recover with enough timeout
re-run of ansible directly after the first exit work fine
Interestingly enough, the skipping works fine with ignore_errors: true when using Mitogen:
TASK [os : Restart network service] ******************************************************************************************************************************************************************************************************
skipping: [host-redacted]
skipping: [host-redacted]
fatal: [host-redacted]: FAILED! => {"msg": "EOF on stream; last 300 bytes received: 'ssh: connect to host aa.bb.cc.dd port 22: Connection refused\\r\\n'"}
...ignoring
This starts to look like a bug to me.