What is the difference between lazarus and zombie killer?

They are just separate services. Lazarus being responsible for restarting, and zombie killer responsible for marking stuff as failed. I don’t think you can look at the services and think why they are split or together because these were open sourced from Cloud but the Cloud versions may still have added functionality

If zombie killer marks a task as failed, will the "failed" status also be communicated/sent to the client flow/task?

No because Zombie Killer marks things as Failed precisely because communication was lost, so it’s a direct change of state on the backend. Does that make sense?

It does. So what happens if the "failed" task actually recovers and completes its unit of work when ZK already marked it failed? Will the existence of a "Done" (on client side) and "Failed" (on cloud) cause inconsistency/conflict?

When that task completes and tries to update state it will run into an error saying the state cant be modified

u rock!

Ah that’s a good thought because I believe it’s Task but Lazarus can also trigger if the Flow never starts (gets stuck waiting for resources)