Handlers are not instance-aware; multiple role invocations in one play restart the wrong recursor or skip a reload

[SrX]

Same class of problem as described for the Authoritative role. The recursor role assumes a single instance per host; invoking it multiple times in one play does not give the handlers per-instance context.

handlers/main.yml (master) restarts via a single listen topic and a flush-time-resolved service name:

- name: Restart pdns-recursor
  ansible.builtin.systemd:
    name: "{{ pdns_rec_service_name }}"
    state: restarted
  listen: restart pdns-recursor
  when: pdns_rec_service_state != 'stopped' and not pdns_rec_disable_handlers

Because handlers are global and keyed by the restart pdns-recursor topic, N notifications from N instances collapse into one run; and because {{ pdns_rec_service_name }} is resolved at flush time (outside the include scope that set it), the single restart does not necessarily target the instance that changed. Net effect: one instance restarts, the others are silently skipped.

Repro sketch: two include_role passes in one play with pdns_rec_service_name: pdns-recursor@a / ...@b, both notifying a restart; end-of-play flush restarts only one. meta: flush_handlers between passes is the current workaround.

Suggested direction: instance-scoped handler names / listen topics, or an explicit single-instance-per-play note in the README.