Unfortunately they don't support shelling out or running static binaries, which makes Cloudflare worthless if you want to run other programming languages like Crystal, Ruby or Python, or do OpenCV stuff, which you can readily do (with some difficulty) in Google Cloud Functions and in Lambda. You are stuck in pure js land with Cloudflare workers, though they are still awesome for this speed increase.
disclaimer: I have 27 Google Cloud Functions doing native stuff atm, about 70% in Crystal, 15% in Ruby, and 15% in Python.
This is a side effect of their execution model. If they had to dedicate one entire OS process to each execution, it would be significantly more expensive. They use V8 Isolates to sandbox the user code, of which you can run many concurrently in a single process cheaply.
Cloudflare Workers are fascinating. I love the idea of distributed functions-as-a-service. Simple deployment, quick response times. Has anyone tried pairing these with a geo-replicated database like Azure's CosmosDB to serve up a REST/GraphQL API?
disclaimer: I have 27 Google Cloud Functions doing native stuff atm, about 70% in Crystal, 15% in Ruby, and 15% in Python.