ZooKeeper Service Discovery Fails to Detect Node Shutdown

case 1:

Issue:
ZooKeeper service discovery fails to detect service instances going offline — Critical

Root cause:
In `registry/zookeeper/src/lib.rs`, the second loop within the `diff()` function queries the wrong map:

As a result, `removed_hosts` remains empty, and consumers continue to send requests to "zombie" nodes after the provider has gone offline.

case 2:

Issue:
Incorrect operator precedence in TLS conditional expression — Critical

Root cause:
`!self.keys.len() != 0` always evaluates to `true`; this causes a panic due to `remove(0)` when certificates are present but keys are missing.

Start in registry/zookeeper/src/lib.rs and inspect the second loop in diff(), tracing which map it queries and how removed_hosts is populated. Reproduce or add coverage for provider shutdown so offline instances are removed from discovery. Then locate the TLS conditional described in case 2 and verify certificate/key combinations no longer reach the reported panic path.

Written by the indexing model from the issue text.