@@ -80,6 +80,8 @@ Search existing issues (`gh issue list --search ...`, both open and closed) for
 - High confidence match: label `duplicate`, comment linking the original issue, and close this one.
 - Low confidence: mention "possibly related to #N" in your triage comment; don't close.
 
+Then look for **similar** issues and PRs as well, whether or not they are duplicates: search both open and closed ones for the symptom, the integration and the component it touches (`gh issue list --state all --search ...` and `gh pr list --state all --search ...`). Read the findings in the best matches — what was diagnosed, what was fixed, and what was ruled out — before forming a hypothesis, and link the relevant ones in your triage comment as "related to #N". An issue that shares a symptom or a trigger is not necessarily the same problem: only call it a duplicate when the cause matches, not just the symptom, and treat a past finding as a lead to confirm against current main, not as a conclusion.
+
 ## 5. Investigate against current main
 
 - Sync the clone first and discard anything left over from a previous run, so you are reading the code you think you are:
 
@@ -18,7 +18,7 @@ Arguments: `queue=<dir> limit=<n>`. Every `*.md` directly in that directory is o
 
 Read it end to end before touching anything. You need to know what is already covered — most candidates restate something the journal says, and the right outcome for those is to drop them, or to sharpen the existing entry rather than add a second one.
 
-Note its own "Adding to this file" section: name the symbol rather than the line number, cite the issue number, keep it short.
+Note its own "What belongs here" and "Adding to this file" sections: keep techniques and traps, one-line device or API facts, and short dated findings that explain a class of report; leave per-issue narrative (what was tried, the fix, rejected alternatives) to the PR and issue. Name the symbol rather than the line number, cite the issue number, keep it short.
 
 ## 2. Verify every candidate against current main
 
@@ -32,7 +32,7 @@ For each candidate, check the claim still holds — grep the symbol it names, re
 
 - **Fold it in** — place it in the right table row or section, rewritten in the journal's voice. Say what was verified and how.
 - **Rewrite it** — the observation is real but the explanation was wrong or has been overtaken.
-- **Drop it** — already covered, unverifiable, or fixed since. Dropping is a normal outcome; say so in the PR body with the reason.
+- **Drop it** — already covered, unverifiable, fixed since, or a per-issue narrative that belongs in the PR or issue rather than the journal. Dropping is a normal outcome; say so in the PR body with the reason.
 
 Anything the candidate flagged as suspected rather than verified stays marked that way, or comes out. Do not upgrade a hypothesis to a fact because it reads well.
 
@@ -46,7 +46,7 @@ git log --oneline --since="2 weeks ago" origin/main
 
 Look for merges touching areas the journal makes claims about. For each, ask whether an existing entry is now wrong: a bug it calls "still live" that has been fixed, a line citation that has drifted onto unrelated code, a `confirmed on main (checked ...)` that a later merge has invalidated. Correct those in the same PR, and say plainly in the body which entries changed and why.
 
-If a fix has landed, say so and keep the mechanism — "this was the bug, it was fixed in PR #N" is useful to a reader holding an older log. Do not simply delete the entry.
+If a fix has landed, say so and keep the mechanism — "this was the bug, it was fixed in PR #N" is useful to a reader holding an older log. Cut the entry down to that rather than leaving the full story, and do not simply delete it. Where the code an entry describes has since changed, correct or delete the entry; do not leave it marked stale.
 
 ## 4. Spelling
 
 
@@ -65,6 +65,7 @@ BLAS
 boostable
 Bottlecap
 brickatius
+britkat1980
 bwignore
 byok
 cadata
 
@@ -1852,6 +1852,7 @@
         "can_span_midnight": True,
         "charge_discharge_with_rate": False,
         "target_soc_used_for_discharge": False,
+        "rate_step_percent_of_capacity": 1,
     },
     "GEC": {
         "name": "GivEnergy Cloud",
 
@@ -33,6 +33,9 @@
 DYNAMIC_LOAD_CAR_SENSOR_MINUTES = 2
 DYNAMIC_LOAD_CAR_START_MINUTES = 3
 DYNAMIC_LOAD_CAR_LOAD_MINUTES = 5
+# A car_charging_now reading only confirms the half hour it is in (keeping that half hour cheap once the car
+# stops, GH#5316) from this many minutes into it: the sensor can lag a car that stopped just before the boundary
+DYNAMIC_LOAD_CAR_CONFIRM_MINUTES = 2
 
 # Extra cloud divergence applied to the PV10 scenario on top of the computed cloud factor, so the
 # downside case diverges harder than the central one
 
@@ -100,6 +100,13 @@ def _clean_number(value):
     return None
 
 
+def _clean_percent(value):
+    """A percentage above 0 and up to 100, or None. Bools are refused, unlike _clean_number."""
+    if isinstance(value, bool) or not isinstance(value, (int, float)):
+        return None
+    return value if 0 < value <= 100 else None
+
+
 def _clean_token(value):
     """A vocabulary token (lower case, no spaces), or None."""
     return value if isinstance(value, str) and VOCAB_RE.match(value) else None
@@ -159,6 +166,7 @@ def _clean_measure_or_tokens(value):
     "min": _clean_number,
     "max": _clean_number,
     "step": _clean_number,
+    "step_percent_of_capacity": _clean_percent,
     "precision": _clean_number,
     "options": _clean_option_list,
     "format": _clean_string,
@@ -1319,6 +1327,12 @@ def not_applicable_field(field):
     else:
         gaps.append("output_charge_control")
         not_applicable_field("current_dp")
+    # charge_rate's step stands for both rates - an inverter holding them coarsely holds both alike. Most
+    # inverters have no such step, so its absence is not a gap: base's value stands, else none
+    if rate is not None and rate.get("step_percent_of_capacity"):
+        definition["rate_step_percent_of_capacity"] = rate["step_percent_of_capacity"]
+    else:
+        definition.setdefault("rate_step_percent_of_capacity", 0)
 
     definition["time_button_press"] = _bound_entity(entities, "schedule_write_button", access="rw") is not None
 
 
@@ -2641,7 +2641,7 @@ def get_car_charging_planned(self):
         self.car_charging_exclusive = [False for c in range(self.num_cars)]
 
         self.car_charging_planned_response = [str(response).lower() for response in self.get_arg("car_charging_planned_response", ["yes", "on", "enable", "true"])]
-        self.car_charging_now_response = [str(response).lower() for response in self.get_arg("car_charging_now_response", ["yes", "on", "enable", "true"])]
+        self.car_charging_now_response = [str(response).lower() for response in self.get_arg("car_charging_now_response", ["yes", "on", "enable", "true", "charging"])]
         self.car_charging_from_battery = self.get_arg("car_charging_from_battery")
 
         # Car charging planned sensor
 
@@ -100,6 +100,11 @@ def _serial_suffix(serial):
 PLAN_MODE_CHARGE = 1
 PLAN_MODE_DISCHARGE = 2
 
+# GivEnergy holds the charge and discharge rates as a whole percent of nominal battery capacity,
+# rounded down - 1300W on a 13.41kWh battery reads back 1206W (#5324)
+GIVENERGY_RATE_STEP_PERCENT_OF_CAPACITY = 1
+GIVENERGY_INVERTER_TYPES = (pb.INVERTER_TYPE_GIVENERGY, pb.INVERTER_TYPE_GIVENERGY_EMS, pb.INVERTER_TYPE_GIVENERGY_GATEWAY)
+
 # Entity attribute table — keyed by the semantic suffix used in dashboard_item calls
 GATEWAY_ATTRIBUTE_TABLE = {
     # Binary sensors
@@ -979,8 +984,15 @@ def _inject_inverter_entities(self, inv, suffix):
         _raw_export_limit = control.export_limit_w
         export_limit_publish = 99999 if _raw_export_limit == 0 else (0 if _raw_export_limit == 1 else _raw_export_limit)
         self.dashboard_item(f"sensor.{pfx}_export_limit_w", export_limit_publish, attributes=GATEWAY_ATTRIBUTE_TABLE.get("export_limit_w", {}), app="gateway")
-        self.dashboard_item(f"number.{pfx}_charge_rate", control.charge_rate_w, attributes=GATEWAY_ATTRIBUTE_TABLE.get("charge_rate", {}), app="gateway")
-        self.dashboard_item(f"number.{pfx}_discharge_rate", control.discharge_rate_w, attributes=GATEWAY_ATTRIBUTE_TABLE.get("discharge_rate", {}), app="gateway")
+        # The hub reports the rate the inverter holds, so a GivEnergy one reads back short of the rate written
+        # (#5324). Marked per inverter, on a copy of the table entry: GWMQTT is one type for every brand
+        charge_rate_attributes = dict(GATEWAY_ATTRIBUTE_TABLE.get("charge_rate", {}))
+        discharge_rate_attributes = dict(GATEWAY_ATTRIBUTE_TABLE.get("discharge_rate", {}))
+        if inv.type in GIVENERGY_INVERTER_TYPES:
+            charge_rate_attributes["step_percent_of_capacity"] = GIVENERGY_RATE_STEP_PERCENT_OF_CAPACITY
+            discharge_rate_attributes["step_percent_of_capacity"] = GIVENERGY_RATE_STEP_PERCENT_OF_CAPACITY
+        self.dashboard_item(f"number.{pfx}_charge_rate", control.charge_rate_w, attributes=charge_rate_attributes, app="gateway")
+        self.dashboard_item(f"number.{pfx}_discharge_rate", control.discharge_rate_w, attributes=discharge_rate_attributes, app="gateway")
         # The reserve ceiling is per-inverter, so it overrides the table's 100: GivEnergy
         # firmware refuses a reserve of 100 and the gateway reports 98 for it (gateway
         # issue #346). adjust_reserve() honours this entity's "max" through
@@ -1239,10 +1251,19 @@ def automatic_config(self):
             # NOTE: control commands are addressed to the Gateway/EMS serial — the firmware
             # must fan these out to the AIOs (tracked separately in command_handler.cpp).
             inverters = gateway_units[:1]
-        else:
+        elif aios:
             inverters = aios
-        if not inverters:
-            inverters = candidate_aios or list(all_inverters)  # last resort
+        else:
+            # A retained startup frame can contain the discovered inverter list before
+            # battery telemetry has been populated.  Treating every discovered unit as
+            # a last-resort control target makes a PV-only inverter writable and the
+            # decision then stays sticky because its serial is not new on later frames.
+            # Keep auto-config incomplete so _needs_reconfigure() retries when the next
+            # status supplies enough capability data to identify a battery inverter.
+            available = [inv.serial for inv in candidate_aios]
+            self.log(f"Warn: GatewayMQTT: no battery-capable inverter telemetry yet (discovered serials: {available}); auto-config deferred")
+            self._auto_configured = False
+            return
 
         # Apply serial filter if configured. A no-match is an error — configuring the
         # wrong inverter set is worse than not configuring at all. Leave _auto_configured
@@ -1291,6 +1312,9 @@ def automatic_config(self):
         discharge_enable_entities = []
         export_limit_entities = []
         inverter_limit_entities = []
+        battery_scaling_entities = []
+        battery_rate_max_entities = []
+        inverter_time_entities = []
 
         for inv in inverters:
             suffix = inv.serial[-6:].lower()
@@ -1323,6 +1347,10 @@ def automatic_config(self):
                 self.log(f"Warn: GatewayMQTT: inverter {inv.serial} has no battery capacity, setting to None for automatic discovery")
 
             inverter_limit_entities.append(f"sensor.{base}_inverter_rate_max")
+            battery_scaling_entities.append(f"sensor.{base}_battery_dod")
+            battery_rate_max_entities.append(f"sensor.{base}_battery_rate_max")
+            # Clock drift detection — uses GatewayStatus.timestamp
+            inverter_time_entities.append(f"sensor.{base}_inverter_time")
 
         # Map entity lists to PredBat args
         self.set_arg("soc_percent", soc_entities)
@@ -1355,13 +1383,11 @@ def automatic_config(self):
 
         # Battery health (first inverter)
         self.set_arg("battery_temperature_history", f"sensor.{base0}_battery_temperature")
-        self.set_arg("battery_scaling", [f"sensor.{base0}_battery_dod"])
-
-        # Battery rate max
-        self.set_arg("battery_rate_max", [f"sensor.{base0}_battery_rate_max"])
 
-        # Inverter time (clock drift detection — uses GatewayStatus.timestamp)
-        self.set_arg("inverter_time", [f"sensor.{base0}_inverter_time"])
+        # Per-inverter: inverter.py reads these at index=self.id, so each needs one entry per inverter
+        self.set_arg("battery_scaling", battery_scaling_entities)
+        self.set_arg("battery_rate_max", battery_rate_max_entities)
+        self.set_arg("inverter_time", inverter_time_entities)
 
         # EMS aggregate entities (GivEnergy EMS only)
         inv0 = inverters[0]
 
@@ -92,6 +92,10 @@
 # restart. Only endpoints not already being managed are probed, and only ever to add them.
 GIVTCP_REDISCOVER_SECONDS = 3600
 
+# GivEnergy holds the charge and discharge rates as a whole percent of nominal battery capacity,
+# rounded down - 1300W on a 13.41kWh battery reads back 1206W (#5324)
+GIVTCP_RATE_STEP_PERCENT_OF_CAPACITY = 1
+
 # control name -> (domain, GivTCPRest write method name, HA entity attributes)
 GIVTCP_CONTROLS = {
     "charge_rate": ("number", "set_charge_rate", {"unit_of_measurement": "W", "device_class": "power", "icon": "mdi:battery-charging", "min": 0, "max": 20000, "step": 100}),
@@ -1038,6 +1042,7 @@ def _discovery_descriptor(self, n, name, domain, access, attrs, max_battery_rate
         elif options:
             descriptor["options"] = options
         if name in ("charge_rate", "discharge_rate"):
+            descriptor["step_percent_of_capacity"] = GIVTCP_RATE_STEP_PERCENT_OF_CAPACITY
             if max_battery_rate:
                 descriptor["max"] = max_battery_rate
             else: