recipes/mimo-*-sparkring: bump sparkring image to dev-20261001-kraken

The sparkring line has newer builds than the pinned 09-28 (plainstatus):
  dev-20260930-spinwait, dev-20261001-statusrows, and
  dev-20261001-kraken (2026-10-01, karmic-kraken base — the active lab GLM
  lineage).

Bump all three MiMo sparkring recipes to the newest:
  ghcr.io/fujitsupolycom/sparkring@sha256:b01442df...b49567
  (tag dev-20261001-kraken-cuda1342-nccl2323-status034, created 10-01)

Digest-verified: the new tag's manifest digest == b01442df...; the old pin
c977a2d2... == dev-20260928-plainstatus. Aligns with the karmic-kraken
adoption; 4x re-bench to validate (operator-approved).

recipes/mimo26-flash-mopd-2x-sparkring/recipe.toml

@@ -20,7 +20,7 @@
 
 engine = "vllm"
 name = "MiMo-V2.6-Flash-MOPD (SparkRing installer image, 2x DAC pair)"
-docker_image = "ghcr.io/fujitsupolycom/sparkring@sha256:c977a2d2efb7ecf9ea856cd0379fdd93a8913f0770f4627a3ae00a084cfaf582"
+docker_image = "ghcr.io/fujitsupolycom/sparkring@sha256:b01442df4e1496bea2f1339feb629eb61dc8175700c82b0e26abbcebd5b49567"
 entrypoint = ["python3", "/opt/sparkring/toolchain/toolchain.py", "serve"]
 # Floating model: the engine resolves the org/model id via the NAS HF cache
 # (refs/main -> latest snapshot; the flat snapshot dir carries dflash/ for the

recipes/mimo26-flash-mopd-4x-sparkring/recipe.toml

@@ -4,9 +4,9 @@
 # sparkring tp4 profile; wiring/cluster facts stay ours.
 #
 # Image (anonymous digests pull):
-#   ghcr.io/fujitsupolycom/sparkring@sha256:c977a2d2...af582
-#   line dev-20260928-plainstatus-cuda1342-nccl2323-status033 =
-#   eugr/spark-vllm-b12x:nightly-20260924 base + the SparkRing fixes we do
+#   ghcr.io/fujitsupolycom/sparkring@sha256:b01442df...b49567
+#   line dev-20261001-kraken-cuda1342-nccl2323-status034 (karmic-kraken
+#   base, the active lab GLM lineage) + the SparkRing fixes we do
 #   NOT have in our nightly:
 #     * B12X selection-cache correction (restarted ranks reuse tuning;
 #       their record: DS4.1 restart 200-225s vs 652-741s),
@@ -43,7 +43,7 @@
 
 engine = "vllm"
 name = "MiMo-V2.6-Flash-MOPD (SparkRing installer image, 4x switched)"
-docker_image = "ghcr.io/fujitsupolycom/sparkring@sha256:c977a2d2efb7ecf9ea856cd0379fdd93a8913f0770f4627a3ae00a084cfaf582"
+docker_image = "ghcr.io/fujitsupolycom/sparkring@sha256:b01442df4e1496bea2f1339feb629eb61dc8175700c82b0e26abbcebd5b49567"
 entrypoint = ["python3", "/opt/sparkring/toolchain/toolchain.py", "serve"]
 # Floating model: the engine resolves the org/model id via the NAS HF cache
 # (refs/main -> latest snapshot; the flat snapshot dir carries dflash/ for the

recipes/mimo26-pro-mopd-8x-sparkring/recipe.toml

@@ -19,7 +19,7 @@
 
 engine = "vllm"
 name = "MiMo-V2.6-Pro-MOPD (SparkRing installer image, 8x switched)"
-docker_image = "ghcr.io/fujitsupolycom/sparkring@sha256:c977a2d2efb7ecf9ea856cd0379fdd93a8913f0770f4627a3ae00a084cfaf582"
+docker_image = "ghcr.io/fujitsupolycom/sparkring@sha256:b01442df4e1496bea2f1339feb629eb61dc8175700c82b0e26abbcebd5b49567"
 entrypoint = ["python3", "/opt/sparkring/toolchain/toolchain.py", "serve"]
 # Floating model: the engine resolves the org/model id via the NAS HF cache
 # (refs/main -> latest snapshot; the snapshot carries dflash/ for the draft