🐛 fix: address edelauna review feedback (Zoo-Code-Org#632)

- Pin actions/cache to SHA per repo convention
- Add VSCode version to e2e cache key
- Remove dead teardown code in zai test interceptor
- Replace setTimeout(100) with deterministic assertion in zai tests
- Change afterAll to afterEach for spy cleanup

Author: HappyLiang12

.github/workflows/e2e.yml

@@ -37,13 +37,13 @@ jobs:
 
             - name: Cache VS Code test binary
               if: github.event_name != 'pull_request' || steps.e2e-marker.outputs.cache-hit != 'true'
-              uses: actions/cache@v4
+              uses: actions/cache@0057852bfaa89a56745cba8c7296529d2fc39830 # v4
               with:
                   path: |
                       apps/vscode-e2e/.vscode-test/
-                  key: vscode-test-${{ runner.os }}-v1
+                  key: vscode-test-${{ runner.os }}-1.100.0-v1
                   restore-keys: |
-                      vscode-test-${{ runner.os }}-
+                      vscode-test-${{ runner.os }}-1.100.0-
 
             - name: Run mocked E2E tests
               id: run-e2e

apps/vscode-e2e/src/suite/providers/zai.test.ts

@@ -76,10 +76,6 @@ function installZAiFetchInterceptor(
 
 	return () => {
 		globalThis.fetch = original
-		const lastCaptured = capturedRequests[capturedRequests.length - 1]
-		if (capture && lastCaptured) {
-			capture.maxTokens = lastCaptured.maxTokens
-		}
 	}
 }
 
@@ -226,9 +222,11 @@ suite("Z.ai GLM provider", function () {
 		})
 
 		await waitUntilCompleted({ api, taskId })
-		// Allow any pending async requests to finish before snapshotting max_tokens
-		await new Promise((resolve) => setTimeout(resolve, 100))
 		const capturedMaxTokens = requestCapture.maxTokens
+		assert.ok(
+			capturedMaxTokens !== undefined,
+			"max_tokens should have been captured by the fetch interceptor before task completion",
+		)
 
 		const completionMessage = messages.find(
 			({ say, text }) => (say === "completion_result" || say === "text") && text?.trim() === "4",
@@ -270,9 +268,11 @@ suite("Z.ai GLM provider", function () {
 		})
 
 		await waitUntilCompleted({ api, taskId })
-		// Allow any pending async requests to finish before snapshotting max_tokens
-		await new Promise((resolve) => setTimeout(resolve, 100))
 		const capturedMaxTokens = requestCapture.maxTokens
+		assert.ok(
+			capturedMaxTokens !== undefined,
+			"max_tokens should have been captured by the fetch interceptor before task completion",
+		)
 
 		const completionMessage = messages.find(
 			({ say, text }) => (say === "completion_result" || say === "text") && text?.trim() === "4",