ci(fleet): retry transient lane failures and fix the self-repin version SIGPIPE

.github/actions/dispatch-suite/action.yaml

@@ -52,8 +52,8 @@ inputs:
   retry-attempts:
     description: >-
       Total dispatch attempts for this suite. Default 1 keeps the historical
-      behavior (a single dispatch, no retry). Set to 2 or more on the heaviest,
-      most race-prone lane so a whole-suite failure re-dispatches a fresh run.
+      behavior (a single dispatch, no retry). Set to 2 or more on a fan-out lane
+      so a whole-suite failure re-dispatches a fresh run.
       This is safe because every scenario-suite run begins with its own
       "reset and seed dev" step, so a re-dispatch starts from a clean slate.
       Only a non-success conclusion (or a run that never became visible)

.github/actions/fleet-repin/action.yaml

@@ -105,7 +105,15 @@ runs:
         install -m 0755 "$TMPDIR/cascade" /usr/local/bin/cascade
         rm -rf "$TMPDIR"
 
-        INSTALLED=$(cascade version 2>/dev/null | head -n 1 | awk '{print $2}')
+        # `cascade version` prints three lines via separate writes. Piping it
+        # straight into `head -n 1` lets head close the pipe after the first
+        # line while the binary is still writing the rest, which delivers
+        # SIGPIPE (exit 141) to the binary and, under `set -o pipefail`, reds
+        # this step intermittently. Capture the output into a variable first,
+        # then take field two of the first line without a pipe, so no producer
+        # can be signalled. This mirrors the setup-cli action's capture pattern.
+        INSTALLED_RAW=$(cascade version 2>/dev/null || true)
+        read -r _ INSTALLED _ <<<"$INSTALLED_RAW" || true
         # Tolerate a leading v in the self-report so the check tracks the release
         # tag rather than a future ldflags formatting choice.
         echo "fleet-repin: installed cascade version ${INSTALLED} (expected ${RC_BARE})"

.github/workflows/fleet-e2e.yaml

@@ -369,6 +369,11 @@ jobs:
           token: ${{ secrets.CASCADE_STATE_TOKEN }}
           cascade_version: ${{ needs.resolve.outputs.cascade_version }}
           cascade_version_sha: ${{ needs.resolve.outputs.cascade_version_sha }}
+          # Allow one whole-suite re-dispatch so a transient runner
+          # non-acquisition, a mid-poll API blip, or a SIGPIPE does not red the
+          # lane on a single unlucky attempt. Safe because each scenario-suite
+          # run self-resets on its first step, so a re-dispatch starts clean.
+          retry-attempts: '2'
 
   # Lane 1, stage 2: dependents of primary (mutate primary's shared external
   # state), so they only start after primary is green. Two repos run together,
@@ -396,6 +401,11 @@ jobs:
           token: ${{ secrets.CASCADE_STATE_TOKEN }}
           cascade_version: ${{ needs.resolve.outputs.cascade_version }}
           cascade_version_sha: ${{ needs.resolve.outputs.cascade_version_sha }}
+          # Allow one whole-suite re-dispatch so a transient runner
+          # non-acquisition, a mid-poll API blip, or a SIGPIPE does not red the
+          # lane on a single unlucky attempt. Safe because each scenario-suite
+          # run self-resets on its first step, so a re-dispatch starts clean.
+          retry-attempts: '2'
 
   # Lane 2: 4env alone. It is the heaviest and most fragile repo, so it runs in
   # its own dedicated job with nothing beside it. Sequenced AFTER the dependents
@@ -424,8 +434,9 @@ jobs:
           # plus chained multi-env hotfix, conflict, rollback, and merge_queue
           # against one live repo). Allow one whole-suite re-dispatch so a live
           # GitHub eventual-consistency blip does not red the lane. Safe because
-          # each suite run self-resets on its first step. Other lanes keep the
-          # default (no retry) so a retry stays a scoped concession here.
+          # each suite run self-resets on its first step. Every fan-out lane
+          # carries the same single re-dispatch so a transient failure on any of
+          # them re-dispatches once rather than redding the whole fleet.
           retry-attempts: '2'
 
   # Lane 3: the light remainder, capped at two repos in flight. Sequenced AFTER
@@ -454,6 +465,11 @@ jobs:
           token: ${{ secrets.CASCADE_STATE_TOKEN }}
           cascade_version: ${{ needs.resolve.outputs.cascade_version }}
           cascade_version_sha: ${{ needs.resolve.outputs.cascade_version_sha }}
+          # Allow one whole-suite re-dispatch so a transient runner
+          # non-acquisition, a mid-poll API blip, or a SIGPIPE does not red the
+          # lane on a single unlucky attempt. Safe because each scenario-suite
+          # run self-resets on its first step, so a re-dispatch starts clean.
+          retry-attempts: '2'
 
   # Fan-in: this job's conclusion is the rc fleet gate. It needs EVERY lane so a
   # green gate means every selected repo passed; auto-promote keys off this