The target URL was requested the first time and/or the source has some issues (e.g. invalid SSL certificate, etc.), or the article text / raw text is very long so that the analysis takes longer than 100ms.
Depending on the load of the system the result is mostly available within the next or the next few requests. In a real-time environment, it isn’t that critical because you’ll get a full result after another or a few requests.
Depending on your use case it might make sense to wait a few seconds until the second request is made to get the results. We recommend waiting – if you do a one-time analysis of URLs e.g. like a classification/categorization of URLs – up to 90 seconds before the 2nd request to secure to get a full response.