Orchestration events are triggered as a "group" and all of them are rollbacked when one of them fails to not leave behind a dirty state (e.g. then stale DHCP entries or TFTP files). This behaviour is intented and one of the core concepts of Orchestration in the Foreman provisioning model.

Personally I see a very low probability, a non-failing approach would lead to a better overall situation as things like "a user can manually..." are not intended to be part of the Foreman routine.

For that specific case: turn off DNS integration alltogether if DNS is not important for your environment.

I agree with Michael, to disable DNS orchestration, just disassociate DNS proxy from you subnet/domain. The possible improvement I see in orchestration would be to pause the provisioning process and let user "retry" A record creation. Today that's is not possible. I'd suggest changing the issue subject asking for "need a way to retry orchestration steps", would that be fair representation?

I believe this is related: https://github.com/theforeman/smart_proxy_dns_infoblox/pull/23 (WIP)

Going to close this one and will create a new ticket once this is confirmed.