diff --git a/zephcore/adapters/sensors/ZephyrEnvSensors.cpp b/zephcore/adapters/sensors/ZephyrEnvSensors.cpp
index 3c97380..946b50e 100644
--- a/zephcore/adapters/sensors/ZephyrEnvSensors.cpp
+++ b/zephcore/adapters/sensors/ZephyrEnvSensors.cpp
@@ -54,8 +54,9 @@ static struct sensor_entry s_sensors[MAX_SENSORS];
 static uint8_t s_count;
 
 /* Is this sensor usable? A node marked zephyr,deferred-init (a part behind a
- * switched rail) is initialised here first. For any other device
- * device_init() answers -EALREADY and this is a device_is_ready() check. */
+ * switched rail) is initialised here first. A device whose boot-time init
+ * failed (an absent part) has its init run once more here; one that came up
+ * answers -EALREADY, so this is a device_is_ready() check. */
 static bool sensor_ready(const struct device *dev)
 {
 	if (dev == NULL) {
diff --git a/zephcore/boards/nrf52840/meshtracker_x1/meshtracker_x1_nrf52840.dts b/zephcore/boards/nrf52840/meshtracker_x1/meshtracker_x1_nrf52840.dts
index f9e07a0..c5b632e 100644
--- a/zephcore/boards/nrf52840/meshtracker_x1/meshtracker_x1_nrf52840.dts
+++ b/zephcore/boards/nrf52840/meshtracker_x1/meshtracker_x1_nrf52840.dts
@@ -319,8 +319,8 @@
 	 * DRV2605_POWER_UP_DELAY_US *before* drv2605_gpio_config() raises en-gpios,
 	 * then reads the status register over I2C immediately after — so the part
 	 * gets no settling time at all between EN rising and being addressed, and
-	 * NACKs. Without the deferral the driver has already run (and failed) by
-	 * the time the app gets a say, and device_init() would answer -EALREADY. */
+	 * NACKs. Without the deferral the driver has already run (and failed) at
+	 * boot, before the app could raise EN and wait. */
 	haptic: drv2605@5a {
 		compatible = "ti,drv2605";
 		reg = <0x5a>;
diff --git a/zephcore/helpers/pm_esp32_wake.c b/zephcore/helpers/pm_esp32_wake.c
index 88d8736..94abf28 100644
--- a/zephcore/helpers/pm_esp32_wake.c
+++ b/zephcore/helpers/pm_esp32_wake.c
@@ -18,17 +18,20 @@
  *   entry (IRQs locked): save the pin's interrupt type and enable, mask its CPU
  *     interrupt, arm a level wake at the pin's active level, keep the pad's
  *     input live through the sleep (sleep-select off).
- *   exit (IRQs already back on, power.c post_ops): disarm, restore the saved
- *     type, clear the status the level mode latched, unmask, and if the pin
- *     is at its active level, re-raise its interrupt through status_w1ts.
+ *   exit (IRQs still locked; the idle thread unlocks them after the kernel
+ *     clock is corrected): disarm, restore the saved type, clear the status
+ *     the level mode latched, unmask, and if the pin is at its active level,
+ *     re-raise its interrupt through status_w1ts.
  *
- * The CPU interrupt stays masked while the pin is in level mode because on
- * this SoC the exit notifiers run after interrupts are re-enabled: a level
- * interrupt on a line the radio holds high until ClearIrqStatus would storm.
- * The re-raise is the missed edge: the level that woke us was reached while
- * the pin was in level mode, so no edge will come. The driver's own ISR then
- * runs the normal callbacks (the radio's dio1_isr, gpio-keys). A duplicate is
- * harmless: both only queue work that re-reads the hardware.
+ * The pin's interrupt stays masked while it is in level mode: a level
+ * interrupt on a line the radio holds high until ClearIrqStatus would storm
+ * if interrupts ran before the type is restored. Since the 2026-10-22 Zephyr
+ * pin they do not (LLD 13 section 14); the mask keeps that true on a Zephyr
+ * version whose post_ops re-enables them. The re-raise is the missed edge: the level that
+ * woke us was reached while the pin was in level mode, so no edge will come.
+ * The driver's own ISR then runs the normal callbacks (the radio's dio1_isr,
+ * gpio-keys) once interrupts are unlocked. A duplicate is harmless: both only
+ * queue work that re-reads the hardware.
  *
  * The ESP32 GPIO wake source itself (ESP_SLEEP_WAKEUP_GPIO) is switched on by
  * `&gpio0 { wakeup-source; }` in boards/common/pm_esp32.overlay.
diff --git a/zephcore/helpers/qspi_probe.c b/zephcore/helpers/qspi_probe.c
index 1155ff9..438ca6e 100644
--- a/zephcore/helpers/qspi_probe.c
+++ b/zephcore/helpers/qspi_probe.c
@@ -67,10 +67,12 @@ static int qspi_rdid_probe(void)
 		return 0;
 	}
 
-	if (!flash->state->initialized) {
-		/* qspi-ext.dtsi defers the flash to first use, and a repeater
-		 * never mounts /ext. Stealing the pins now would break the init
-		 * a later factory reset runs, so there is nothing to probe. */
+	if (!flash->state->initialized && flash->state->init_res == 0) {
+		/* Init never ran: qspi-ext.dtsi defers the flash to first use,
+		 * and a repeater never mounts /ext. Stealing the pins now would
+		 * break the init a later factory reset runs, so there is nothing
+		 * to probe. A failed init leaves initialized clear too, with the
+		 * errno in init_res (kernel/device.c). */
 		LOG_INF("qspi flash '%s' not initialised yet - RDID probe skipped",
 			flash->name);
 		return 0;
diff --git a/zephcore/west.yml b/zephcore/west.yml
index 985f962..f9d9237 100644
--- a/zephcore/west.yml
+++ b/zephcore/west.yml
@@ -6,7 +6,7 @@ manifest:
   projects:
     - name: zephyr
       remote: zephyrproject-rtos
-      revision: c136c73d7be3305d2c07ee5726d4c6d0d985068b
+      revision: b70dbaf045e06ee721316987df71c35d5a718693
       import: true
 
   self: