Request coalescing (or collapsing) is the CDN behavior of merging many concurrent requests for the same uncached object into one fetch toward the origin, then serving all the waiting clients from that single response. It is the mechanism that prevents a thundering herd from stampeding the origin.

It matters most during synchronized demand - a live premiere or a kickoff - when a brand-new segment is requested by huge numbers of viewers within the same second. Combined with origin shielding, coalescing turns what would be thousands of identical origin pulls into one, protecting origin capacity and keeping delivery stable under spikes.