diff --git a/.github/scripts/shard-tests.sh b/.github/scripts/shard-tests.sh
index 13db2f598..765e3345d 100755
--- a/.github/scripts/shard-tests.sh
+++ b/.github/scripts/shard-tests.sh
@@ -148,6 +148,44 @@ while [ "$i" -lt "$total" ]; do
   i=$((i + 1))
 done
 
+# What a pinned file is seeded with, and why it is not its @test count.
+#
+# The pin exists because these files' @test count does not describe their cost
+# (see above). Seeding `load` with that same count therefore hands the balancer
+# the one number it was told not to trust: a 630s file that contains 9 @test
+# blocks reports a load of 9, the shard reads as very nearly empty, and the
+# ordinary weighted pass fills it right back up. Measured on the terminal-driver
+# tree (macOS, run 33850939213 attempt 4): the four shards came out
+# 32.2/17.1/10.8/12.2 minutes against a 30-minute cap, and the 32.2 was the
+# shard holding the 630s pin plus a full share of everything else. It timed out
+# four attempts in a row while two other shards sat idle for fifteen minutes.
+#
+# So a pinned file is seeded at ONE SHARD'S AVERAGE SHARE instead: it is treated
+# as already worth about what a shard is supposed to hold, which is the honest
+# reading of "we cannot weigh this one". Same tree, same measured per-file
+# seconds: 14.5/15.8/21.0/21.1, a 21-minute worst case.
+#
+# This is deliberately not a table of per-file seconds. That alternative was
+# considered when the weights were first written and rejected because it goes
+# stale silently -- it would be wrong the moment a fixed `sleep` becomes a poll
+# loop, with nothing to say so. The share is derived from the tree on every run
+# and cannot drift away from it.
+#
+# The failure mode to watch is over-reservation: a pinned file that is actually
+# light leaves its shard underfilled and pushes the others up. That is bounded
+# while the pins are few relative to `total` -- at 2 pins in 4 shards the worst
+# case is the remaining work over two shards, which the numbers above clear with
+# room. Pinning a majority of the shards would invert this, and nothing here
+# detects that; keep PINNED_APART short.
+total_tests=0
+while IFS= read -r f; do
+  [ -n "$f" ] || continue
+  total_tests=$((total_tests + $(file_weight "$f")))
+done <<EOF
+$files
+EOF
+pin_seed=$((total_tests / total))
+
 pinned_paths=" "
 slot=0
 for p in $PINNED_APART; do
@@ -163,7 +201,7 @@ $files
 EOF
   [ -n "$match" ] || continue
   s=$((slot % total))
-  load[s]=$((load[s] + $(file_weight "$match")))
+  load[s]=$((load[s] + pin_seed))
   if [ "$s" -eq "$((index - 1))" ]; then
     printf '%s\n' "$match"
   fi
diff --git a/tests/test_ci_sharding.bats b/tests/test_ci_sharding.bats
index 98bf510b0..2cac8754e 100644
--- a/tests/test_ci_sharding.bats
+++ b/tests/test_ci_sharding.bats
@@ -179,3 +179,40 @@ union_of_shards() {
   # log claims a split the workflow is not performing.
   grep -q "bats (\${{ matrix.os }} \${{ matrix.shard }}/$total)" "$wf"
 }
+
+# The pin's whole point is that these files' @test count does not describe their
+# cost. Seeding the balancer with that count undoes the pin: the shard reads as
+# empty and gets refilled, which is how one shard reached 32 minutes against a
+# 30-minute cap while two others sat at 11 and 12 (macOS, run 33850939213).
+#
+# What this asserts is the reservation itself: a shard holding a pinned file
+# must end up carrying MEASURABLY FEWER tests than the average shard, because
+# the seed already spent that shard's budget. A test that only checked the
+# pinned files land on different shards -- which the suite above already does --
+# stays green through exactly this regression, since they still land apart when
+# seeded by count. That is the break this one is here to catch.
+@test "a pinned file's shard is reserved, not refilled (pin seed is not the count)" {
+  local script="$BATS_TEST_DIRNAME/../.github/scripts/shard-tests.sh"
+  local dir="$BATS_TEST_DIRNAME/.."
+  local total=4 n f count mean grand=0
+
+  for f in "$BATS_TEST_DIRNAME"/*.bats; do
+    grand=$((grand + $(grep -c '^[[:space:]]*@test' "$f" || true)))
+  done
+  mean=$((grand / total))
+  [ "$mean" -gt 0 ]
+
+  # Shard 1 and shard 2 hold the two pinned files (slot 0 and slot 1).
+  for n in 1 2; do
+    count=0
+    while IFS= read -r f; do
+      [ -n "$f" ] || continue
+      count=$((count + $(grep -c '^[[:space:]]*@test' "$dir/$f" || true)))
+    done < <(cd "$dir" && bash "$script" "$n" "$total")
+
+    # Strictly below the average, with margin: seeded by count these shards come
+    # out at roughly the mean, so a bare "<= mean" would be a coin flip rather
+    # than a check. Two thirds is well clear of both states.
+    [ "$count" -lt $((mean * 2 / 3)) ]
+  done
+}