Cloudflare runs its main blog on pre-1.0 EmDash, load-tested to 7,000 requests per second

The load-test result depends on how much traffic the caches absorb. The report does not say what share of the 7,000 requests per second in the load test got past Workers Cache and the KV object cache to PlanetScale, or what the hit rate was ^[2][4].

Measured against the blog's own traffic, the test is about 93 times typical load ^[1] but only 1.4 times the level its spikes exceed ^[2]. The second ratio is the one a blog plans around, because spikes are the load it has to survive. For the 7,000 figure to transfer to another site, that site's pages would need to be read far more often than they change, and served mostly from cache.

The latency evidence is a p95 chart, green for the old platform and yellow for EmDash ^[7]. Its authors are Kody Jackson, senior manager of content engineering, systems engineer Diogo Carneiro and senior design engineer Amy Dutton ^[14]. "Where the previous platform experienced periodic latency spikes under load, the new system maintains a remarkably flat, consistent response profile," they wrote ^[8]. They credit the result to "running EmDash on Cloudflare Workers alongside our new caching layers" ^[9]. By that account the chart compares two whole architectures, so it cannot isolate what EmDash contributes apart from the caches.

The cutover is the best-engineered part of the write-up. "This Worker set a version cookie on requests, which then let us route incoming traffic to the new or legacy experience accordingly," the team wrote. "Additionally, this strategy allowed us to fall back to the legacy blog if the new site experienced any 500 errors." ^[11] When EmDash failed with a 500, the reader got the WordPress page. The trigger is narrow, though. A page that renders wrongly with a 200 status passes straight through, and so does a stale cache entry. The fallback also required WordPress to stay up for the whole rollout.

Cloudflare calls itself "Customer Zero" for EmDash ^[6]. Customer Zero has one advantage no outside adopter gets: the CMS is developed in-house, so the editors' bug reports went to colleagues ^[1][13]. Scheduling, one of the problem areas, is also among the actions the new EmDash MCP server gives authors, along with browsing, creating, editing and publishing ^[12][13].