AppSync metrics show <50 ms latency while my client's network latency is showing 200-350ms.

My client is a React app using @aws-amplify/api running on my local machine. In the browser's dev console "Network" tab, I see graphql queries that take 200-350ms. "Waiting for server response" accounts for >99% of the latency.

In AppSync (us-east-2), I have a lambda function data source that runs <8-10 ms consistently and off of cold starts <60 ms. But even when setting the data source to NoneDatasource (the request data is forwarded immediately to subscribers) I see no difference in the problematic 200-350ms latency. Weirdly, I see the AppSync metrics show that I'm getting a respectable <50 ms latency. I'm not using caching.

Was really hoping AppSync would be a solution for real time data but 350ms is far too high. Is it possible I'm missing something (maybe overhead from the aws-amplify package or from using API key for auth?) or is this normal/expected latency?