I've been looking at an old issue, RESTEASY-1709 "Same JAX-RS Provider instance injected across applications", which I thought I had solved by injecting a javassist created proxy, but the reporter couldn't get it to work so it's been sitting for almost a year. Anyway, I've been trying to port the old code to master, and I see that injection has changed to do a lot of stuff with CompletableFutures. In particular, a series of injections into, say, a resource, becomes a composition of CompletionStages, each one doing a single injection. I'm trying to understand why. One reason I can see is that, at least in the case of an asynchronous invocation, you want to do as little as possible before returning a CompletableFuture. That would provide better responsiveness. But I have two questions:   1. Is constructing a CompletableFuture that much faster than just doing the injection?   2. I don't how much space a CompletableFuture takes, but we've turned a simple loop into a collection of CompletableFutures, which would seem to occupy more space. Isn't that counterproductive in terms of our general mandate to create a smaller footprint for container applications? I'm not saying that the new code is wrong. I'm just saying that I don't understand why it's right. Any thoughts? -- My company's smarter than your company (unless you work for Red Hat)