diff --git a/docs/tutorials/snapshots.rst b/docs/tutorials/snapshots.rst
index f17d55d7e..1070aa373 100644
--- a/docs/tutorials/snapshots.rst
+++ b/docs/tutorials/snapshots.rst
@@ -31,9 +31,32 @@ configuration always overrides the zero-config fallback when present.
 
 .. note::
 
-   By default, snapshots are deleted when a fault is cleared via the
-   ``DELETE /api/v1/faults/{code}`` endpoint or ``~/clear_fault`` service.
-   Set ``snapshots.retain_on_clear: true`` to keep them across clears.
+   By default, clearing a fault deletes its **value snapshots**, the per-topic
+   JSON captures this tutorial configures. Every clear reaches storage the same
+   way, whatever asked for it: the per-fault
+   ``DELETE /api/v1/apps/{app_id}/faults/{fault_code}`` route (``components``,
+   ``areas`` and ``functions`` carry the same one, though there is no global
+   ``DELETE /api/v1/faults/{code}``), the bulk
+   ``DELETE /api/v1/apps/{app_id}/faults`` and ``DELETE /api/v1/faults``, the
+   ``~/clear_fault`` service, and the correlation cascade, which clears a root
+   cause's symptoms with no clear addressed to them. The OPC UA plugin calls
+   that same service when a threshold alarm de-asserts, so a value going back
+   in range clears the fault the way an operator does; a native AlarmCondition
+   clears that way only once the operator has acknowledged it (and confirmed
+   it, unless ``require_confirm_for_clear`` is off), and stays CONFIRMED until
+   then. A plugin that de-asserts with a PASSED report instead, as the graph
+   watchdog does, goes through debounce and healing, and nothing on that path
+   deletes the snapshots: with ``healing_enabled`` the fault heals once its
+   counter reaches ``healing_threshold``, without it (the default) it stays
+   CONFIRMED, and its snapshots stay either way. Starting with healing disabled takes the
+   value snapshots of leftover HEALED rows as it reclassifies them, without a
+   clear at all. ``snapshots.retain_on_clear: true`` keeps the value snapshots
+   across all of it, and only those. The rosbag recording is not covered by that
+   setting. A clear deletes the recording under
+   ``snapshots.rosbag.auto_cleanup``, unless that is off or
+   ``snapshots.rosbag.max_bags_per_fault`` keeps a history. The cap then governs
+   retention, except at ``0`` (unlimited), where only
+   ``snapshots.rosbag.max_total_storage_mb`` bounds it.
 
 Quick Start
 -----------
@@ -360,7 +383,7 @@ Troubleshooting
 
 **Empty topics object in response**
 
-- The fault may have been cleared (snapshots are deleted on clear unless
+- The fault may have been cleared (value snapshots are deleted on clear unless
   ``snapshots.retain_on_clear`` is enabled)
 - No topics were configured for this fault code
 - All configured topics timed out or exceeded size limit
@@ -380,7 +403,10 @@ Rosbag Capture (Time-Window Recording)
 
 In addition to JSON snapshots, you can enable **rosbag capture** for "black box"
 style recording. This continuously buffers messages in memory and flushes them
-to a bag file when a fault is confirmed.
+to a bag file when a fault is confirmed. The recording is deleted when the fault
+is cleared, unless ``snapshots.rosbag.auto_cleanup`` is off or
+``snapshots.rosbag.max_bags_per_fault`` keeps a history, which takes over the
+retention decision.
 
 **Key differences from JSON snapshots:**
 
diff --git a/src/ros2_medkit_fault_manager/README.md b/src/ros2_medkit_fault_manager/README.md
index c12e54c8a..977f5a8c4 100644
--- a/src/ros2_medkit_fault_manager/README.md
+++ b/src/ros2_medkit_fault_manager/README.md
@@ -52,7 +52,7 @@ ros2 service call /fault_manager/clear_fault ros2_medkit_msgs/srv/ClearFault \
 - **Severity escalation**: Fault severity is updated if a higher severity is reported
 - **Persistent storage**: SQLite backend ensures faults survive node restarts
 - **Debounce filtering** (optional): AUTOSAR DEM-style counter-based fault confirmation with per-entity threshold overrides
-- **Snapshot capture**: Captures topic data when faults are confirmed for debugging (snapshots are deleted when fault is cleared)
+- **Snapshot capture**: Captures topic data when faults are confirmed for debugging (the value snapshots are deleted when the fault is cleared, unless `snapshots.retain_on_clear` is set)
 - **Near-miss series**: Appends one entry per FAILED report that moved the debounce counter without confirming, bounded per fault code and retained when the fault is cleared
 - **Freeze-frame retention**: One compact JSON freeze-frame per fault code, retained across `clear_fault` (see below)
 - **Fault correlation** (optional): Root cause analysis with symptom muting and auto-clear
diff --git a/src/ros2_medkit_fault_manager/config/snapshots.yaml b/src/ros2_medkit_fault_manager/config/snapshots.yaml
index 6e7179f20..a4efadfed 100644
--- a/src/ros2_medkit_fault_manager/config/snapshots.yaml
+++ b/src/ros2_medkit_fault_manager/config/snapshots.yaml
@@ -192,6 +192,8 @@ rosbag:
   # Auto-cleanup bag files when fault is cleared (default: true)
   # When true, bag files are deleted when ClearFault is called
   # When false, bag files persist until manually deleted or storage limit hit
+  # No effect once max_bags_per_fault is anything other than 1: a configured
+  # history governs retention instead, and an acknowledgement leaves it alone
   auto_cleanup: true
 
 # =============================================================================
diff --git a/src/ros2_medkit_gateway/README.md b/src/ros2_medkit_gateway/README.md
index 0cff2b0d3..5c0f55990 100644
--- a/src/ros2_medkit_gateway/README.md
+++ b/src/ros2_medkit_gateway/README.md
@@ -1267,7 +1267,7 @@ Rosbag capture is configured via FaultManager parameters. See `config/snapshots.
 | `snapshots.rosbag.exclude_sensor_topics` | bool | `true` | Auto-exclude image/points/depth/compressed in broad modes (`include_topics` re-adds) |
 | `snapshots.rosbag.qos_match` | bool | `true` | Match each topic's publisher QoS for faithful capture |
 | `snapshots.rosbag.format` | string | `"mcap"` | Bag format: `"mcap"` (default; opens directly in Foxglove and Lichtblick) or `"sqlite3"`. Neither is privileged - an unavailable plugin falls back automatically to the other one |
-| `snapshots.rosbag.auto_cleanup` | bool | `true` | Delete bag when fault is cleared |
+| `snapshots.rosbag.auto_cleanup` | bool | `true` | Delete a fault's bag when the fault is cleared. No effect once `snapshots.rosbag.max_bags_per_fault` is anything other than `1`, where the cap governs retention instead |
 | `snapshots.rosbag.max_buffer_mb` | int | `256` | Ring-buffer RAM cap (oldest messages drop past it) |
 | `snapshots.rosbag.max_bag_size_mb` | int | `50` | Max size per bag file |
 | `snapshots.rosbag.max_total_storage_mb` | int | `500` | Total storage limit |
diff --git a/src/ros2_medkit_plugins/ros2_medkit_opcua/README.md b/src/ros2_medkit_plugins/ros2_medkit_opcua/README.md
index 0381ef765..bf262d56e 100644
--- a/src/ros2_medkit_plugins/ros2_medkit_opcua/README.md
+++ b/src/ros2_medkit_plugins/ros2_medkit_opcua/README.md
@@ -875,7 +875,14 @@ GET /api/v1/apps/tank_process/faults
   -> [{fault_code: "PLC_HIGH_TEMP", severity: "ERROR", ...}]
 ```
 
-When the value returns below threshold, the fault is automatically cleared.
+When the value returns below threshold, the plugin calls the fault manager's
+`~/clear_fault` service (`/fault_manager/clear_fault`) for that fault code. That
+is the same service an operator's
+`DELETE /api/v1/apps/{app_id}/faults/{fault_code}` ends up calling, so a device
+de-assert is a clear like any other. It drops the fault's value snapshots unless
+`snapshots.retain_on_clear` is set, and its rosbag recording unless
+`snapshots.rosbag.auto_cleanup` is off or `snapshots.rosbag.max_bags_per_fault`
+keeps a history.
 
 ## Key Design Decisions