diff --git a/.claude/skills/sister-project-history/otopcua.md b/.claude/skills/sister-project-history/otopcua.md
index 25466db..4e80c04 100644
--- a/.claude/skills/sister-project-history/otopcua.md
+++ b/.claude/skills/sister-project-history/otopcua.md
@@ -221,3 +221,42 @@ AddRoute of the dev Mac (`oee-tagdump-mac`, `10.181.206.245.1.1`) on the two tes
     - On SqlServer/QuestDB nodes, `Total` aggregate values change meaning.
     - Event HistoryRead may now return continuation points, which a client must follow.
     - Otherwise none by default.
+- **2026-10-12 follow-up batch #674–#678 (OtOpcUa `6d5c1c76`).** #674, #676 and #678 are closed. #675 and #677 stay
+  open only for live AVEVA checks that need the VPN. New issues: #679 (Gateway `Count`/`Average` mode mapping) and
+  #680 (pre-existing forward differences from the SDK aggregate calculator). Three docker-dev live gates passed; one
+  failure (A1) was fixed and re-gated.
+  - **#674 (HistoryUpdate):**
+    - **Primary-only:** a non-Primary node returns `BadInvalidState`. The gate stays closed until the node's first
+      redundancy snapshot.
+    - **Future skew:** a timestamp beyond `now + ServerHistorian:HistoryUpdateMaxFutureSkew` (default 1 day) is
+      rejected with `BadOutOfRange`.
+    - **SqlServer v1 schema:** v1 is accepted again. On v1, modified reads and HistoryUpdate return
+      `BadHistoryOperationUnsupported`, but historization continues.
+    - **QuestDB TTL:** confirmed to follow the wall clock, so a far-future row cannot wipe history. The issue's
+      premise was false.
+  - **#675 (events):** reverse event HistoryRead pages newest-first on SqlServer and QuestDB. Gateway refuses an
+    explicit reverse read. An end-only capped read goes backward where the backend supports it and stays forward on
+    Gateway.
+  - **#677 (aggregates):**
+    - ⚠️ Reverse processed reads now follow the SDK / Part 13: anchored at the later time, partial bucket earliest,
+      later-edge timestamps. Forward reads are unchanged.
+    - A start-of-data Partial bit is set on forward reads.
+    - A non-default client `AggregateConfiguration` is rejected with `BadAggregateConfigurationRejected`.
+    - Gateway `IntegralDivisor` is now 1.
+  - **#676 (OpcUaClient):**
+    - Fast recovery: a definitive session loss re-initialises immediately instead of waiting for the ~90 s watchdog.
+      Definitive means a session status, `BadServerHalted`, a sustained closed secure channel, or a subscription the
+      server dropped. This is exposed through the new `IDriverConnectionLossSource`.
+    - Restart recovery went from ~122 s to 12–27 s. Pause blips still recover in place.
+    - The driver now rewrites the advertised endpoint's host and port.
+    - A trailing re-read repairs values lost in flight.
+    - The `ReconnectPeriod` setting is removed.
+  - **#678 (clients and role overlays):**
+    - Client.CLI/UI now rewrite the endpoint port, which fixes the rig `:4841` "certificate" error.
+    - ⚠️ Per-role appsettings overlays never loaded on cluster-role nodes; they now do. Cluster-role nodes therefore
+      pick up `Security.Ldap.Transport=Ldaps`, so set `Security__Ldap__Transport` explicitly on any plaintext-LDAP
+      central node.
+  - **ScadaBridge impact:**
+    - OpcUaClient-backed tags recover much faster after an upstream restart.
+    - Only reverse reads change shape (ScadaBridge issues no processed or event history reads).
+    - HistoryUpdate is refused off-Primary.