While reviewing a GTmetrix performance report, I noticed unusually high server response times despite having an A-grade score. The HAR file revealed that all resources were being served over HTTP/1.1 instead of HTTP/2, even though I believed HTTP/2 had been enabled previously. I queried my CloudFront distribution using the AWS CLI and confirmed the HttpVersion setting was indeed set to http1.1. I updated the distribution configuration to use HTTP/2, which enables multiplexing (allowing multiple requests over a single connection) and binary header compression. After the change propagated to CloudFront's edge locations, I ran a second GTmetrix test. The results were striking: Largest Contentful Paint dropped from 690ms to 167ms, Time to First Byte collapsed from 386ms to under 5ms, and the fully loaded time fell from 2 seconds to just 193ms. The improvement came from two factors working together: HTTP/2's multiplexing eliminated connection overhead, and the warmed CloudFront cache meant all resources were served directly from the edge location without origin fetches.