While debugging, I had a look at the bundle that Rollup generates inside the Karma test runner. It appears to wrap the test module in some code that injects the setImmediate function from Node.js. This is probably done by @rollup/plugin-node-resolve. If this is indeed the case, then the tests are not actually checking whether fastTimeout is implemented correctly, but whether setImmediate is a good alternative. This could be determined by running the tests in debug mode and setting a breakpoint in the fastTimeout function. If the breakpoint is never encountered, we are testing setImmediate instead.

A possible fix is to sophisticate the export default statement in src/fastTimeout.js. Right now, it prioritizes setImmediate if present. This could be changed to prioritize our own fastTimeout if postMessage is present. This might be a good idea, anyway. There should probably also be a third alternative in case neither is present; _.defer would work for that scenario, which is equivalent to _.partial(setTimeout, _, 0).