diff --git a/.tekton/acceptance-tests-pr.yaml b/.tekton/acceptance-tests-pr.yaml
index 1a25ab9e..c74de833 100644
--- a/.tekton/acceptance-tests-pr.yaml
+++ b/.tekton/acceptance-tests-pr.yaml
@@ -93,3 +93,18 @@ spec:
             requests:
               storage: 200Mi
           storageClassName: standard
+    # ── Go module cache ──────────────────────────────────────────────────────
+    # Shared persistent volume populated once by the warm-go-cache Task before
+    # parallel release-tests Tasks start.  All test pods mount this PVC and
+    # find all Go dependencies already present, eliminating the concurrent
+    # module downloads that caused gauge runner timeouts (>101 min).
+    #
+    # Pre-requisite (apply once per cluster):
+    #   oc apply -f tekton/pvc/go-module-cache.yaml -n openshift-pipelines
+    #
+    # The plumbing acceptance-tests Pipeline and release-tests Task also need
+    # to be updated to accept and use this workspace (see
+    # tekton/tasks/warm-go-cache.yaml and the companion plumbing PR).
+    - name: go-cache
+      persistentVolumeClaim:
+        claimName: go-module-cache
diff --git a/.tekton/create-ci-image-pipelinerun.yaml b/.tekton/create-ci-image-pipelinerun.yaml
index ff5bac15..6e6eea62 100644
--- a/.tekton/create-ci-image-pipelinerun.yaml
+++ b/.tekton/create-ci-image-pipelinerun.yaml
@@ -5,7 +5,7 @@ metadata:
   name: create-push-ci-image
   annotations:
     pipelinesascode.tekton.dev/on-cel-expression: |
-      "Dockerfile.CI".pathChanged() && ( event == "push" || event == "pull_request" )
+      ( "Dockerfile.CI".pathChanged() || "go.mod".pathChanged() || "go.sum".pathChanged() ) && ( event == "push" || event == "pull_request" )
     pipelinesascode.tekton.dev/max-keep-runs: "5"
 spec:
   taskRunSpecs:
diff --git a/Dockerfile.CI b/Dockerfile.CI
index 421bdd55..411b00f4 100644
--- a/Dockerfile.CI
+++ b/Dockerfile.CI
@@ -62,6 +62,23 @@ RUN wget https://github.com/getgauge/gauge/releases/download/v${GAUGE_VERSION}/g
     go env -w GOPROXY="https://proxy.golang.org,direct" &&\
     gauge version
 
+# ── Go module cache pre-population ────────────────────────────────────────────
+# When multiple release-tests tasks run in parallel across OCP cluster versions,
+# each task previously created a fresh empty GOPATH and downloaded all Go
+# modules independently, causing network saturation and gauge runner timeouts
+# (>101 min).  By pre-downloading the full module graph here, every container
+# spawned from this image already has all deps available at /go/pkg/mod without
+# any network access.  GOMODCACHE is set as an ENV so it is inherited by every
+# task process, keeping it independent of the per-task temporary GOPATH.
+# The CI image is rebuilt by PAC whenever Dockerfile.CI, go.mod or go.sum
+# changes, so the baked cache is always consistent with the module graph.
+ENV GOMODCACHE=/go/pkg/mod
+COPY go.mod go.sum /go-cache-warmup/
+RUN cd /go-cache-warmup && \
+    go env -w GOMODCACHE=/go/pkg/mod && \
+    go mod download -x && \
+    rm -rf /go-cache-warmup
+
 RUN wget https://github.com/sigstore/cosign/releases/download/v3.0.3/cosign-linux-amd64 -O /usr/bin/cosign && \
     chmod a+x /usr/bin/cosign
 
diff --git a/tekton/pvc/go-module-cache.yaml b/tekton/pvc/go-module-cache.yaml
new file mode 100644
index 00000000..dd04bc8f
--- /dev/null
+++ b/tekton/pvc/go-module-cache.yaml
@@ -0,0 +1,55 @@
+# go-module-cache PVC
+# ─────────────────────────────────────────────────────────────────────────────
+# Shared, persistent PVC for the Go module download cache.
+#
+# Why ReadWriteMany?
+#   Multiple release-tests TaskRun pods run in parallel (one per test suite).
+#   All of them need to mount the same volume concurrently - that requires an
+#   RWX access mode.  On ROSA/OSD/OCP the default StorageClass is usually
+#   backed by Ceph RBD (RWO only).  Ensure you use a StorageClass that supports
+#   ReadWriteMany (e.g. ocs-storagecluster-cephfs, nfs, or azurefile).
+#
+# Lifecycle:
+#   - This PVC is long-lived (not recreated per PipelineRun).
+#   - It is populated once by the warm-go-cache Task before parallel test tasks
+#     start.  Subsequent runs reuse the existing cache; only changed modules are
+#     fetched.
+#   - If go.mod / go.sum change significantly, delete and recreate the PVC to
+#     avoid stale entries (or let warm-go-cache handle it with go mod download,
+#     which is additive-only).
+#
+# Apply once per cluster with:
+#   oc apply -f tekton/pvc/go-module-cache.yaml -n openshift-pipelines
+# ─────────────────────────────────────────────────────────────────────────────
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: go-module-cache
+  namespace: openshift-pipelines
+  labels:
+    app.kubernetes.io/part-of: release-tests
+    app.kubernetes.io/component: go-cache
+  annotations:
+    # Human-readable explanation of why this PVC exists
+    release-tests.tekton.dev/purpose: >
+      Shared Go module cache populated once per pipeline run by the
+      warm-go-cache task, then mounted read-only by all parallel
+      release-tests tasks to eliminate concurrent module downloads that
+      cause gauge runner timeout (>101 min).
+spec:
+  accessModes:
+    # ReadWriteMany is required so all parallel test pods can mount
+    # simultaneously.  Use ocs-storagecluster-cephfs or equivalent.
+    - ReadWriteMany
+  resources:
+    requests:
+      # The full Go module graph for this project is ~1.5 GB.
+      # 5 Gi gives comfortable headroom for future growth.
+      storage: 5Gi
+  # Replace with the RWX-capable StorageClass available on your cluster.
+  # Examples:
+  #   OCP/OSD/ROSA (OCS)   : ocs-storagecluster-cephfs
+  #   ARO                  : azurefile-csi
+  #   GKE                  : standard-rwx
+  #   Generic NFS          : nfs-client
+  storageClassName: ocs-storagecluster-cephfs
diff --git a/tekton/stepactions/warm-go-cache-with-oci.yaml b/tekton/stepactions/warm-go-cache-with-oci.yaml
new file mode 100644
index 00000000..e7dde8dc
--- /dev/null
+++ b/tekton/stepactions/warm-go-cache-with-oci.yaml
@@ -0,0 +1,145 @@
+# tekton-caches integration for cross-run Go module caching
+# ─────────────────────────────────────────────────────────────────────────────
+# This file enhances the warm-go-cache Task (tekton/tasks/warm-go-cache.yaml)
+# with tekton-caches StepActions so that the Go module cache is also persisted
+# across pipeline runs in an OCI registry.
+#
+# How it works (two-tier strategy):
+#
+#   Tier 1 – within a single pipeline run (PVC):
+#     warm-go-cache Task runs once → fills PVC → all parallel test tasks
+#     mount the PVC and find modules locally.
+#
+#   Tier 2 – across pipeline runs (OCI registry via tekton-caches):
+#     cache-fetch StepAction pulls the tarball from Quay.io at the start of
+#     warm-go-cache; if the hash matches, the PVC is pre-filled in seconds
+#     (no module-proxy download at all).
+#     cache-upload StepAction pushes a fresh tarball when the hash changes.
+#
+# Cache key: SHA-256 hash of go.sum (content-addressed → auto-invalidates on
+# any dependency change, no manual purge needed).
+#
+# Pre-requisites:
+#   1. StepActions feature gate must be enabled in TektonCD Pipelines:
+#        kubectl patch configmap -n tekton-pipelines --type merge \
+#          -p '{"data":{"enable-step-actions":"true"}}' feature-flags
+#   2. A robot account on Quay.io with write access to
+#      quay.io/openshift-pipeline/release-tests-go-cache.
+#      Create a Tekton secret and link it to the SA used by the pipeline.
+#
+# Apply with:
+#   kubectl apply -f tekton/stepactions/cache-fetch.yaml
+#   kubectl apply -f tekton/stepactions/cache-upload.yaml
+#   (or use the upstream versions directly from tekton-caches)
+# ─────────────────────────────────────────────────────────────────────────────
+
+# ── Updated warm-go-cache Task with tekton-caches integration ─────────────────
+apiVersion: tekton.dev/v1
+kind: Task
+metadata:
+  name: warm-go-cache-with-oci
+  annotations:
+    tekton.dev/tags: "go,cache,oci"
+    tekton.dev/displayName: "Warm Go Module Cache (OCI-backed)"
+    tekton.dev/description: >
+      Populates the shared go-module-cache PVC using a two-tier strategy:
+      first tries to restore from OCI registry via tekton-caches (fast, cross-run),
+      then falls back to go mod download if the cache is stale or absent.
+      After a fresh download the new cache is uploaded back to the registry.
+spec:
+  params:
+    - name: IMAGE
+      description: CI image (Go + gauge)
+      default: quay.io/openshift-pipeline/ci
+    - name: REGISTRY
+      description: >
+        OCI registry path for the cache image, with {{hash}} placeholder.
+        Example: oci://quay.io/openshift-pipeline/release-tests-go-cache:{{hash}}
+      default: "oci://quay.io/openshift-pipeline/release-tests-go-cache:{{hash}}"
+    - name: GOPROXY
+      description: Go module proxy URL
+      default: "https://proxy.golang.org,direct"
+  workspaces:
+    - name: release-tests-git
+      description: Checked-out release-tests repository (go.mod / go.sum)
+    - name: go-cache
+      description: Shared RWX PVC for Go module cache (GOMODCACHE)
+    - name: dockerconfig
+      description: >
+        Workspace containing Docker config for OCI registry auth.
+        Bind to a secret holding .dockerconfigjson for Quay.io.
+      optional: true
+  steps:
+    # ── Step 1: restore cache from OCI registry ────────────────────────────
+    - name: cache-fetch
+      ref:
+        # Upstream StepAction from tekton-caches project.
+        # Pin to a specific digest in production.
+        resolver: http
+        params:
+          - name: url
+            value: https://raw.githubusercontent.com/openshift-pipelines/tekton-caches/main/tekton/cache-fetch.yaml
+      params:
+        - name: PATTERNS
+          value: ["go.sum"]
+        - name: SOURCE
+          value: $(params.REGISTRY)
+        - name: CACHE_PATH
+          value: $(workspaces.go-cache.path)/gopath/pkg/mod
+        - name: WORKING_DIR
+          value: $(workspaces.release-tests-git.path)
+        - name: DOCKER_CONFIG
+          value: $(workspaces.dockerconfig.path)
+
+    # ── Step 2: download any missing modules from module proxy ─────────────
+    - name: download-modules
+      image: $(params.IMAGE)
+      imagePullPolicy: Always
+      workingDir: $(workspaces.release-tests-git.path)
+      env:
+        - name: GOPATH
+          value: $(workspaces.go-cache.path)/gopath
+        - name: GOMODCACHE
+          value: $(workspaces.go-cache.path)/gopath/pkg/mod
+        - name: GOPROXY
+          value: $(params.GOPROXY)
+        - name: CACHE_FETCHED
+          value: $(steps.cache-fetch.results.fetched)
+      script: |
+        #!/usr/bin/env bash
+        set -eu -o pipefail
+
+        mkdir -p "${GOMODCACHE}"
+
+        if [[ "${CACHE_FETCHED}" == "true" ]]; then
+          echo "==> OCI cache hit – verifying completeness ..."
+          # Quick sanity check: if go mod download exits 0 with no output it
+          # means all deps are present; any missing module is downloaded.
+          go mod download 2>&1 | head -50
+        else
+          echo "==> OCI cache miss – downloading full module graph ..."
+          go mod download -x
+        fi
+
+        echo "==> Module cache size: $(du -sh ${GOMODCACHE} | cut -f1)"
+
+    # ── Step 3: upload refreshed cache to OCI registry ────────────────────
+    - name: cache-upload
+      ref:
+        resolver: http
+        params:
+          - name: url
+            value: https://raw.githubusercontent.com/openshift-pipelines/tekton-caches/main/tekton/cache-upload.yaml
+      params:
+        - name: PATTERNS
+          value: ["go.sum"]
+        - name: TARGET
+          value: $(params.REGISTRY)
+        - name: CACHE_PATH
+          value: $(workspaces.go-cache.path)/gopath/pkg/mod
+        - name: WORKING_DIR
+          value: $(workspaces.release-tests-git.path)
+        - name: FETCHED
+          value: $(steps.cache-fetch.results.fetched)
+        - name: DOCKER_CONFIG
+          value: $(workspaces.dockerconfig.path)
diff --git a/tekton/tasks/warm-go-cache.yaml b/tekton/tasks/warm-go-cache.yaml
new file mode 100644
index 00000000..0d7c03ff
--- /dev/null
+++ b/tekton/tasks/warm-go-cache.yaml
@@ -0,0 +1,90 @@
+# warm-go-cache Task
+# ─────────────────────────────────────────────────────────────────────────────
+# Runs ONCE before all parallel release-tests Tasks in the acceptance-tests
+# Pipeline.  It downloads the full Go module graph for this project into the
+# shared go-module-cache PVC so that every concurrent test pod finds its
+# dependencies locally and performs zero network downloads.
+#
+# Placement in the acceptance-tests Pipeline (plumbing repo):
+#   runAfter: [clone-release-tests-git]
+#   (all parallel release-tests tasks get runAfter: [warm-go-cache])
+#
+# This Task must be added to plumbing/ci/tasks/warm-go-cache.yaml and
+# referenced from plumbing/ci/pipelines/acceptance-tests.yaml.
+# ─────────────────────────────────────────────────────────────────────────────
+apiVersion: tekton.dev/v1
+kind: Task
+metadata:
+  name: warm-go-cache
+  annotations:
+    tekton.dev/tags: "go,cache"
+    tekton.dev/displayName: "Warm Go Module Cache"
+    tekton.dev/description: >
+      Populates the shared go-module-cache PVC by running go mod download
+      once before parallel test tasks start.  Each test task can then set
+      GOPATH to the PVC mount path and skip all module downloads, avoiding
+      the gauge runner timeout caused by concurrent network downloads.
+spec:
+  params:
+    - name: IMAGE
+      description: CI image that has Go, gauge and the same GOPROXY setting
+      default: quay.io/openshift-pipeline/ci
+    - name: GOPROXY
+      description: Go module proxy URL
+      default: "https://proxy.golang.org,direct"
+  workspaces:
+    - name: release-tests-git
+      description: Checked-out release-tests repository containing go.mod/go.sum
+    - name: go-cache
+      description: >
+        Shared PVC that will hold $GOPATH/pkg/mod after this task completes.
+        Must be bound to the go-module-cache ReadWriteMany PVC.
+  steps:
+    - name: download-modules
+      image: $(params.IMAGE)
+      imagePullPolicy: Always
+      workingDir: $(workspaces.release-tests-git.path)
+      env:
+        # Point GOPATH and GOMODCACHE at the shared PVC so that
+        # go mod download writes modules to persistent storage.
+        - name: GOPATH
+          value: $(workspaces.go-cache.path)/gopath
+        - name: GOMODCACHE
+          value: $(workspaces.go-cache.path)/gopath/pkg/mod
+        - name: GOPROXY
+          value: $(params.GOPROXY)
+        # Disable sum-db verification for internal/air-gapped environments
+        - name: GONOSUMCHECK
+          value: "*"
+      script: |
+        #!/usr/bin/env bash
+        set -eu -o pipefail
+
+        echo "==> Go module cache warm-up"
+        echo "    GOPATH      : ${GOPATH}"
+        echo "    GOMODCACHE  : ${GOMODCACHE}"
+        echo "    GOPROXY     : ${GOPROXY}"
+        echo "    go.mod hash : $(sha256sum go.mod | cut -d' ' -f1)"
+        echo "    go.sum hash : $(sha256sum go.sum | cut -d' ' -f1)"
+
+        # Create the module cache directory if it doesn't exist yet
+        mkdir -p "${GOMODCACHE}"
+
+        # Check whether the cache is already current by comparing the
+        # go.sum hash stored from the previous run.
+        HASH_FILE="${GOMODCACHE}/.go-sum-hash"
+        CURRENT_HASH=$(sha256sum go.sum | cut -d' ' -f1)
+
+        if [[ -f "${HASH_FILE}" ]] && [[ "$(cat ${HASH_FILE})" == "${CURRENT_HASH}" ]]; then
+          echo "==> Cache is up-to-date (go.sum unchanged). Skipping download."
+          exit 0
+        fi
+
+        echo "==> Downloading Go modules (this populates the shared PVC) ..."
+        go mod download -x
+
+        # Persist the go.sum hash so future runs can skip the download
+        echo -n "${CURRENT_HASH}" > "${HASH_FILE}"
+
+        echo "==> Done. Module cache size:"
+        du -sh "${GOMODCACHE}" || true
\ No newline at end of file