diff --git a/.github/required-checks.txt b/.github/required-checks.txt
index 3128ff2..535ee75 100644
--- a/.github/required-checks.txt
+++ b/.github/required-checks.txt
@@ -10,3 +10,4 @@
 
 lint
 unit (L1)
+vm-e2e
diff --git a/.github/workflows/auto-release.yml b/.github/workflows/auto-release.yml
index f04e89e..ed6dea7 100644
--- a/.github/workflows/auto-release.yml
+++ b/.github/workflows/auto-release.yml
@@ -173,16 +173,16 @@ jobs:
           **Why:** ${REASON}
 
           Because this release includes \`feat:\` changes, the sensor did
-          **not** auto-tag. Run the destructive e2e suite locally, then
-          cut the release manually:
+          **not** auto-tag. Confirm the merge and post-merge checks, then cut
+          the release manually:
 
           - [ ] L4 CI (\`vm-e2e-spike.yml\`) is green on the latest commit on \`main\`
           - [ ] sanity-check the curl|bash smoke and cli-compat results in the most recent test.yml run on main
           - [ ] \`git tag -a ${NEW_TAG} -m "..."\` and \`git push origin ${NEW_TAG}\`
           - [ ] close this issue
 
-          L4 CI is not yet a hard merge gate, but \`feat:\` changes carry
-          more risk than \`fix:\` patches — verify it before tagging.
+          L4 CI is a required merge check; verify the latest \`main\` run too
+          because \`feat:\` changes carry more risk than \`fix:\` patches.
           EOF
           )
 
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index 37330ce..5e77e54 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -65,6 +65,7 @@ jobs:
         run: make test-unit
 
   vm-e2e:
+    name: release vm-e2e (L4)
     needs: gate-tests
     runs-on: macos-14
     timeout-minutes: 60
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index 3ad8652..997eaca 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -37,7 +37,7 @@ Tests are split across four tiers. Which one runs where:
 | **L1 Unit + Integration + Contract** | Pure-Go logic with faked `Runner` *plus* real `brew` / `git` / `npm` against temp dirs and real `httptest` servers | `make test-unit` (~75s) | Every push (pre-push hook); CI on push/PR |
 | **L2 Contract schema** | JSON schema validation against [openboot-contract](https://github.com/openbootdotdev/openboot-contract) | (runs in CI only) | CI on push/PR |
 | **L3 E2E binary** | Compiled binary driven by scripts; `-tags=e2e` | `make test-e2e` | CI on release |
-| **L4 VM e2e** | Full destructive suite (`-tags="e2e,vm"`). Installs real packages, modifies `~/.zshrc`, writes `defaults`. Each run requires a clean macOS host (Apple Silicon). | CI only — `vm-e2e-spike.yml` on `macos-14` | **CI** — GitHub Actions `macos-14` runner, two parallel jobs. No local target. |
+| **L4 VM e2e** | Full destructive suite (`-tags="e2e,vm"`). Installs real packages, modifies `~/.zshrc`, writes `defaults`. Each run requires a clean macOS host (Apple Silicon). | CI only — `vm-e2e-spike.yml` on `macos-14` | Every PR as the required `vm-e2e` check. No local target. |
 
 Rules of thumb:
 
diff --git a/docs/HARNESS.md b/docs/HARNESS.md
index 8caa48c..4b84bfb 100644
--- a/docs/HARNESS.md
+++ b/docs/HARNESS.md
@@ -51,9 +51,10 @@ Three regulation categories:
 | Behav. | L1 unit + integration + contract (faked runners *and* real brew/git/npm in temp dirs) | pre-push, CI | `make test-unit` |
 | Behav. | L2 contract schema (against openboot-contract repo) | CI | `.github/workflows/test.yml` `contract` job |
 | Behav. | L3 e2e binary | release | `make test-e2e` |
-| Behav. | L4 VM e2e (`vm`) — full destructive suite on a clean macOS host | every PR | `.github/workflows/vm-e2e-spike.yml` (macos-14 runner, two parallel jobs) |
+| Behav. | L4 VM e2e (`vm`) — required full destructive suite on a clean macOS host | every PR | `.github/workflows/vm-e2e-spike.yml` (`vm-e2e` required check on a macos-14 runner) |
+| Behav. | `install.sh` upgrade over an existing install — tap refresh, upgrade, reinstall fallback, and resolved-version reporting with fake Homebrew/OpenBoot commands | L1 | `test/integration/install_script_test.go` |
 | Behav. | Install-wizard TUI on a real pty — L3: launch/quit smoke + full keyboard choreography (stops before confirm, installs nothing); L4: same key sequence through a real install via `expect(1)`, asserting brew/git system state | L3 at release, L4 every PR | `test/e2e/install_wizard_e2e_test.go`, `test/e2e/install_wizard_vm_test.go` |
-| Behav. | curl\|bash smoke (install.sh + mock server) | every PR | `.github/workflows/test.yml` `curl-bash-smoke` job |
+| Behav. | curl\|bash smoke (install.sh + mock server) | push to main / dispatch | `.github/workflows/test.yml` `curl-bash-smoke` job |
 | Behav. | Auto-release sensor — patch fast lane (`fix:`-only) auto-tags + dispatches `release.yml`; feat threshold opens a `release-ready` issue (check L4 CI green, then tag manually) | push to `main` | `.github/workflows/auto-release.yml` |
 | Behav. | Release notes — Conventional Commits since previous tag, grouped by type (Features / Bug Fixes / etc) + Full Changelog link, appended to the install-instructions template | tag push or `workflow_dispatch` | `.github/workflows/release.yml` (`Write release notes` step) |
 | Behav. | Old-CLI compat (previous release × current mock server) | every PR | `.github/workflows/test.yml` `cli-compat` job |
@@ -76,7 +77,7 @@ When you observe a recurring issue, decide where to encode the fix:
 | "Agent doesn't know about preset X." | Update `internal/config/data/presets.yaml`. Source of truth, not docs. |
 | "Agent introduced a new lint failure that golangci-lint should have caught." | Enable the relevant linter in `.golangci.yml`. |
 | "Agent broke a behaviour that has no test." | Write the test at the right tier — L1 covers both faked-runner units in `internal/<pkg>/` and real-subprocess integration in `test/integration/`. |
-| "A shipped release reached nobody: `install.sh`'s already-installed branch prompted on stdin, which under `curl \| bash` is the script itself, so it always took the don't-upgrade default." | Already handled by the `installsh` archtest. The wider lesson the sensor does *not* cover: `curl-bash-smoke` is gated `if: github.event_name != 'pull_request'` and only exercises the mock-server path, so the real `scripts/install.sh` brew branch has no behavioural test. Upgrade-over-existing-install is the case to add. |
+| "A shipped release reached nobody: `install.sh`'s already-installed branch prompted on stdin, which under `curl \| bash` is the script itself, so it always took the don't-upgrade default." | The `installsh` archtest blocks stdin prompts, and `TestIntegration_InstallScript_ExistingInstallUpgrade` drives the real script through piped stdin with fake commands to pin tap refresh, upgrade, reinstall fallback, and version reporting. |
 | "Agent missed an AGENTS.md rule we keep restating." | Make it a hard or soft archtest rule (a docs rule that doesn't fail is a docs rule that drifts). |
 | "Agent did something safe but suboptimal." | Add to AGENTS.md "Project-specific conventions" and consider whether it's encodable. |
 | "Agent guessed at an API contract." | Update `openboot-contract` repo + fixtures; CI already runs schema validation. |
@@ -119,9 +120,8 @@ it survives doc rot.
 - **L4 runs on GitHub Actions, not a self-hosted runner.** `macos-14`
   runners are Apple Silicon VMs — each job gets a fresh clean macOS
   environment, which is exactly what L4 needs. Tart is no longer required.
-  The L4 workflow (`vm-e2e-spike.yml`) is not yet a hard merge gate (not in
-  `required-checks.txt`); it runs on every PR. Promoting it to a required
-  check is the next step once the workflow has proven stable.
+  The L4 workflow (`vm-e2e-spike.yml`) runs on every PR as the required
+  `vm-e2e` merge check.
 
 ## How agents should think about this file
 
diff --git a/docs/MERGE_POLICY.md b/docs/MERGE_POLICY.md
index 3e09096..71e2fcb 100644
--- a/docs/MERGE_POLICY.md
+++ b/docs/MERGE_POLICY.md
@@ -21,6 +21,7 @@ Runs on every PR. Must pass before merge.
 |---|---|---|
 | `lint` | Test | Catches gofmt / gosec / staticcheck issues that block release builds. |
 | `unit (L1)` | Test | Unit + integration + contract: faked-runner Go tests *and* real `brew` / `git` / `npm` against temp dirs. Includes `internal/archtest` fitness rules. |
+| `vm-e2e` | vm-e2e-spike | Exercises the destructive install paths and TUI choreography on a fresh Apple Silicon macOS VM. |
 
 ### Post-merge (runs on push to `main`)
 
@@ -38,7 +39,6 @@ the auto-release sensor can tag. Runs on `workflow_dispatch` and
 
 | Check | Status | Reason |
 |---|---|---|
-| `macos e2e (L4)` | runs only on tag pushes / manual dispatch | Slow + destructive; runs at release time, not per PR. |
 | Harness drift sensors (`govulncheck`, `deadcode`, `mod-tidy diff`, `archtest stale baseline`) | `continue-on-error: true` | Informational by design. Failures surface as annotations and, on `main`, open tracking issues via `drift-to-issue.yml`. |
 | `codecov/patch` | informational | Coverage threshold is a guideline, not a gate. Hard coverage gates push toward test-shaped code without raising actual quality. |
 
@@ -56,7 +56,7 @@ the auto-release sensor can tag. Runs on `workflow_dispatch` and
   add it to this list. Promote a check to required by editing this doc
   and updating branch protection in the same PR.
 
-## Why these two
+## Why these three
 
 Each required check covers a class of regression that has shipped to
 users in past commits:
@@ -65,6 +65,9 @@ users in past commits:
 - `unit (L1)` is the broadest behaviour check — covers both faked-runner
   unit logic and real-subprocess integration drift (brew flag changes,
   `git` exit-code shifts between macOS versions).
+- `vm-e2e` (L4) covers the destructive and terminal-dependent paths that
+  cannot safely run inside L1, including real Homebrew installs and the
+  install-wizard choreography on a fresh macOS VM.
 
 The three heavier checks (`contract schema (L2)`, `curl|bash smoke`,
 `old-cli compat`) still run on every merge to `main` — they just don't
diff --git a/test/integration/install_script_test.go b/test/integration/install_script_test.go
new file mode 100644
index 0000000..397d3af
--- /dev/null
+++ b/test/integration/install_script_test.go
@@ -0,0 +1,115 @@
+package integration
+
+import (
+	"bytes"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestIntegration_InstallScript_ExistingInstallUpgrade(t *testing.T) {
+	script, err := os.ReadFile(filepath.Join("..", "..", "scripts", "install.sh"))
+	require.NoError(t, err)
+
+	tests := []struct {
+		name        string
+		upgradeExit string
+		wantCalls   []string
+	}{
+		{
+			name:        "upgrades after refreshing the tap",
+			upgradeExit: "0",
+			wantCalls: []string{
+				"brew list openboot",
+				"brew update",
+				"brew upgrade openbootdotdev/tap/openboot",
+				"openboot version",
+				"openboot install --help",
+			},
+		},
+		{
+			name:        "reinstalls when upgrade fails",
+			upgradeExit: "1",
+			wantCalls: []string{
+				"brew list openboot",
+				"brew update",
+				"brew upgrade openbootdotdev/tap/openboot",
+				"brew reinstall openbootdotdev/tap/openboot",
+				"openboot version",
+				"openboot install --help",
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			tmpDir := t.TempDir()
+			fakeBin := filepath.Join(tmpDir, "bin")
+			require.NoError(t, os.Mkdir(fakeBin, 0o755))
+
+			callLog := filepath.Join(tmpDir, "calls.log")
+			writeFakeCommand(t, fakeBin, "uname", `#!/bin/bash
+case "${1:-}" in
+    -s) echo Darwin ;;
+    -m) echo arm64 ;;
+    *) exit 90 ;;
+esac
+`)
+			writeFakeCommand(t, fakeBin, "xcode-select", "#!/bin/bash\nexit 0\n")
+			writeFakeCommand(t, fakeBin, "brew", `#!/bin/bash
+printf 'brew %s\n' "$*" >> "$CALL_LOG"
+case "${1:-}" in
+    list)
+        [[ "${2:-}" == "openboot" ]] || exit 91
+        exit 0
+        ;;
+    update) exit 0 ;;
+    upgrade) exit "${BREW_UPGRADE_EXIT:-0}" ;;
+    reinstall) exit 0 ;;
+    *) exit 92 ;;
+esac
+`)
+			writeFakeCommand(t, fakeBin, "openboot", `#!/bin/bash
+printf 'openboot %s\n' "$*" >> "$CALL_LOG"
+case "${1:-}" in
+    version) echo "openboot version v-test" ;;
+    install) exit 0 ;;
+    *) exit 93 ;;
+esac
+`)
+
+			// Feed the script through bash's stdin to exercise the documented
+			// `curl | bash` execution mode as well as the already-installed path.
+			cmd := exec.Command("/bin/bash", "-s", "--", "--help")
+			cmd.Stdin = bytes.NewReader(script)
+			cmd.Env = []string{
+				"PATH=" + fakeBin + ":/usr/bin:/bin",
+				"HOME=" + filepath.Join(tmpDir, "home"),
+				"CALL_LOG=" + callLog,
+				"BREW_UPGRADE_EXIT=" + tt.upgradeExit,
+			}
+
+			output, err := cmd.CombinedOutput()
+			require.NoError(t, err, "install.sh output:\n%s", output)
+
+			calls, err := os.ReadFile(callLog)
+			require.NoError(t, err)
+			assert.Equal(t, tt.wantCalls, strings.Split(strings.TrimSpace(string(calls)), "\n"))
+			assert.Contains(t, string(output), "OpenBoot is already installed — updating...")
+			assert.Contains(t, string(output), "✓ OpenBoot updated!")
+			assert.Contains(t, string(output), "openboot version v-test")
+			assert.NotContains(t, string(output), "Reinstall?")
+			assert.NotContains(t, string(output), "Installing OpenBoot via Homebrew...")
+		})
+	}
+}
+
+func writeFakeCommand(t *testing.T, dir, name, contents string) {
+	t.Helper()
+	require.NoError(t, os.WriteFile(filepath.Join(dir, name), []byte(contents), 0o755))
+}