Has Azure Storage ever permanently lost data protected by ZRS?

I don't have a confirmed example of permanent data loss from ZRS, or a reliable figure for the longest outage. A regional outage can make data inaccessible without destroying it, so those are separate questions when you're reviewing incident reports.

For the plan, consider both the cost of losing data and the cost of being unable to use it. Cross-region replication adds cost, but it addresses a different failure scenario from redundancy within one region. Historical incidents alone won't tell you whether that protection is worth it for your workload.

Replication isn't a backup. An accidental deletion or bad change can affect replicated data too. Keep a separate backup with a recovery process you can test, rather than treating ZRS or geo-redundancy as the only protection against permanent loss.

Check durability and availability separately in Microsoft's documentation. Durability concerns whether the data survives. Availability concerns whether you can access it when needed. A high durability figure doesn't promise that a regional outage will be short.

For OneLake, also check the current scope and recovery behavior of the optional cross-region feature. Replicated files and folders by themselves don't establish that all Fabric infrastructure and workloads can fail over.

The 2018 South Central US outage is a useful case to study for recovery decisions, but it shouldn't be counted as an example of ZRS data loss. It predates ZRS, and the account discussed involved concern about potential loss during failover rather than confirmed permanent data loss.