If not, what would be a workaround? It would be optimal, if the Fly.io Prometheus instance would remain the main source of metrics, for such a scenario to work.
Generally the answer to " Is there a way to adjust Fly.io’s Prometheus configuration to provide valid data for a self-hosted AlertManager instance?" is “No”. As you’re aware, alerting rules are set on Prometheus, not on Alertmanager, and there’s no way to set custom rules on Fly Prometheus.