docs(capm3): add Unhealthy Annotation section and update SUMMARY

docs/user-guide/src/SUMMARY.md

@@ -23,7 +23,7 @@
       - [RAID Setup](bmo/raid.md)
       - [Rebooting Hosts](bmo/reboot_annotation.md)
       - [Specifying Root Device](bmo/root_device_hints.md)
-   - [Advanced Features](bmo/advanced-features.md)
+  - [Advanced Features](bmo/advanced-features.md)
       - [Adopting Externally Provisioned Hosts](bmo/externally_provisioned.md)
       - [Advanced Instance Customization](bmo/advanced_instance_customization.md)
       - [Booting from Live ISO](bmo/live-iso.md)

docs/user-guide/src/capm3/introduction.md

@@ -16,6 +16,64 @@ itself is shared across multiple cloud providers. Cluster API Provider Metal3 is
 one of the providers for Cluster API and enables users to deploy a Cluster API based
 cluster on top of bare metal infrastructure using Metal3.
 
+
+## Unhealthy Annotation
+
+The `capi.metal3.io/unhealthy` annotation is used by Cluster API Provider Metal³ (CAPM3)
+to mark **BareMetalHost** objects that should not be selected for provisioning new
+`Metal3Machine` resources.
+
+When this annotation is present, CAPM3 excludes the annotated host from consideration
+when matching available hardware to new Machines. This prevents the reuse of hosts that
+are known to be unhealthy or have failed remediation attempts.
+
+## Manual usage
+
+Operators can manually mark a host as unhealthy by adding the following annotation to a
+`BareMetalHost` object:
+
+```yaml
+metadata:
+  annotations:
+    capi.metal3.io/unhealthy: "true"
+```
+
+Removing the annotation re-enables the host for normal provisioning by CAPM3.
+
+## Automatic application after remediation timeout
+
+Starting from CAPM3 API version `v1alpha4` (available in previous release branches),
+this annotation may also be **applied automatically** when remediation attempts time
+out and the node fails to recover.
+
+During a remediation cycle managed by a `Metal3Remediation` resource, the following
+parameters define retry and timeout behavior:
+
+- `.spec.strategy.retryLimit` — the number of reboot retries permitted before the
+  remediation is considered failed.
+- `.spec.strategy.timeout` — the duration to wait between retries for the node to
+  become healthy.
+
+If the final timeout expires and the node remains unhealthy:
+
+1. CAPM3 sets the `MachineOwnerRemediatedCondition=False` condition on the affected
+   `Machine` to begin deletion of the unhealthy `Machine` and related remediation
+   objects.
+2. The corresponding `BareMetalHost` is automatically annotated with:
+
+```yaml
+   metadata:
+     annotations:
+       capi.metal3.io/unhealthy: "true"
+```
+
+This automatic annotation ensures that CAPM3 does not immediately attempt to reuse the
+same physical host for another Machine after remediation failure. The host remains
+excluded from new provisioning until an operator manually removes the annotation after
+verifying and correcting the underlying issue.
+
+Check the [Remediation](https://book.metal3.io/capm3/remediaton/) process fro more details.
+
 ## Compatibility with Cluster API
 
 | CAPM3 version | Cluster API version | CAPM3 Release |