Review other (non-netobserv) metrics available out there, and see if we can leverage our alerting+health mechanism on them too

E.g:

• ingress errors (haproxy_server_http_responses_total)
• ingress performance degrading (? haproxy_server_http_average_response_latency_milliseconds)
• ingress connections coming close to capacity
• apiserver errors (code:apiserver_request_total:rate5m{apiserver="kube-apiserver"})
• apiserver tls handshake errors (cluster:apiserver_tls_handshake_errors_total:rate5m{apiserver="kube-apiserver"})
• apiserver performance degrading (??)
• ovn error (ovnkube_node_cni_request_duration_seconds_count{err!="false"})

See also these RFEs: RFE-2935, RFE-8004 

The latter provides some examples of promQL to alert on. Also, it's an interesting point to rely also on node-exporter metrics as they don't overlap with netobserv own drop metrics (kernel drops vs device drops)