[AustinBenoit]

Summary

Reduces flakiness in FirebaseAppCheckTest.TestGetTokenForcingRefresh:

1. Use GetAppCheckToken(false) for the initial token request (GetToken #1):
   On Apple platforms (iOS / tvOS), FIRAppCheck is eagerly instantiated during App::Create() with isTokenAutoRefreshEnabled = true and AppCheckCoreTokenRefresher immediately schedules an initial background periodicTokenRefresh (token(forcingRefresh: false)). Calling GetAppCheckToken(true) (forcingRefresh: true) immediately afterward bypassed AppCheckCore's ongoingTask coalescing and launched a competing token refresh request that raced on Keychain storage, occasionally overwriting the newly fetched token with the slightly older background token before GetToken #2 (force_refresh=false) ran. Calling GetAppCheckToken(false) for GetToken #1 coalesces with any in-flight startup token fetch.
2. Increase sleep intervals between token fetches from 500ms to 1000ms:
   Ensures distinct expiration timestamps across all platforms (including Desktop where response.fetch_time() has 1-second granularity).

[gemini-code-assist]

Code Review

This pull request modifies the TestGetTokenForcingRefresh integration test by changing the first token request to not force a refresh and increasing the sleep durations from 500ms to 1000ms. The review comments suggest updating a misleading comment at line 515 to correctly reflect that the cached token is expected, and adding a clarifying comment at line 525 to explain the purpose of the sleep before forcing a token refresh.

[gemini-code-assist]

The comment here states that we wait to make sure the expire time would be different, but the subsequent assertion on line 522 actually expects the expire times to be equal (EXPECT_EQ). This comment is misleading for this specific step and should be updated to reflect that we are verifying the cached token is still returned.

Suggested change

	// Wait a bit to make sure the expire time would be different
	std::this_thread::sleep_for(std::chrono::milliseconds(500));
	std::this_thread::sleep_for(std::chrono::milliseconds(1000));
	// Wait a bit to verify the cached token is still returned
	std::this_thread::sleep_for(std::chrono::milliseconds(1000));

[gemini-code-assist]

This sleep is where we actually need to ensure that the clock has advanced so that the newly forced token gets a different expiration timestamp (especially on desktop platforms with 1-second granularity). Adding a clarifying comment here would improve readability.

Suggested change

	std::this_thread::sleep_for(std::chrono::milliseconds(1000));
	// Wait a bit to make sure the new token's expire time will be different
	std::this_thread::sleep_for(std::chrono::milliseconds(1000));