AVVideoComposition startRequest early return without calling finishWithComposedVideoFrame

@AVFoundationEngineers I browsed through AVCustomEdit sample code and notice the following:

- (void)startVideoCompositionRequest:(AVAsynchronousVideoCompositionRequest *)request

{
	@autoreleasepool {
		dispatch_async(_renderingQueue,^() {
             // Check if all pending requests have been cancelled
			if (_shouldCancelAllRequests) {
				[request finishCancelledRequest];
			} else {
				NSError *err = nil;
				// Get the next rendererd pixel buffer
				CVPixelBufferRef resultPixels = [self newRenderedPixelBufferForRequest:request error:&err];

				if (resultPixels) {
					// The resulting pixelbuffer from OpenGL renderer is passed along to the request
					[request finishWithComposedVideoFrame:resultPixels];
					CFRelease(resultPixels);
				} else {
					[request finishWithError:err];

				}
			}
       }
}

The startVideoComposition request returns early without calling finishWithComposedVideoFrame as the processing takes place asynchronously in a dispatchQueue. However, the documentation of startVideoCompositionRequest states the following:

	Note that if the custom compositor's implementation of -startVideoCompositionRequest: returns without finishing the composition immediately, it may be invoked again with another composition request before the prior request is finished; therefore in such cases the custom compositor should be prepared to manage multiple composition requests.

But I don't see anything in the code that is prepared to handle multiple composition requests of same video frame. How is one supposed to handle this case?