There are a few traffic `check_prometheus` based alerts that should be migrated to Alertmanager:
```
modules/monitoring/manifests/alerts/aggregate_ipsec.pp: monitoring::check_prometheus { "aggregate-ipsec-tunnel-status-${site}":
modules/monitoring/manifests/alerts/http_availability.pp: monitoring::check_prometheus { "varnish_${title}":
modules/monitoring/manifests/alerts/http_availability.pp: monitoring::check_prometheus { "ats_${title}":
modules/monitoring/manifests/alerts/traffic_drop.pp: monitoring::check_prometheus { $title:
modules/profile/manifests/cache/kafka/varnishkafka_delivery_alert.pp: monitoring::check_prometheus { "varnishkafka-${instance}-${cache_segment}-eqiad-kafka_drerr":
modules/profile/manifests/cache/kafka/varnishkafka_delivery_alert.pp: monitoring::check_prometheus { "varnishkafka-${instance}-${cache_segment}-codfw-kafka_drerr":
modules/profile/manifests/cache/kafka/varnishkafka_delivery_alert.pp: monitoring::check_prometheus { "varnishkafka-${instance}-${cache_segment}-esams-kafka_drerr":
modules/profile/manifests/cache/kafka/varnishkafka_delivery_alert.pp: monitoring::check_prometheus { "varnishkafka-${instance}-${cache_segment}-ulsfo-kafka_drerr":
modules/profile/manifests/cache/kafka/varnishkafka_delivery_alert.pp: monitoring::check_prometheus { "varnishkafka-${instance}-${cache_segment}-eqsin-kafka_drerr":
modules/profile/manifests/cache/purge.pp: monitoring::check_prometheus { 'purged-event-lag':
modules/profile/manifests/cache/purge.pp: monitoring::check_prometheus { 'purged-backlog':
modules/profile/manifests/cache/varnish/frontend.pp: monitoring::check_prometheus { 'varnishd-mmap-count':
modules/profile/manifests/lvs.pp: monitoring::check_prometheus { 'excessive-lvs-rx-traffic':
modules/profile/manifests/lvs.pp: monitoring::check_prometheus { 'lvs-cpu-saturated':
modules/pybal/manifests/monitoring.pp: monitoring::check_prometheus { 'pybal_bgp_sessions':
modules/varnish/manifests/instance.pp: monitoring::check_prometheus { 'varnish-frontend-check-child-start':
```
Notably the "reduced availability" alerts can suffer from delays or excessive averaging due to their use of the prometheus `global` instance and icinga evaluation period.