Known differences

As of 1.x release

Topics such as Cross-Origin, Content Security Policy, Mixed Content, Service Workers are ignored, given our server-side context.
URL input must be an absolute URL, using either http or https as scheme.
On the upside, there are no forbidden headers, and res.url contains the final url when following redirects.
For convenience, res.body is a transform stream, so decoding can be handled independently.
Similarly, req.body can either be a string, a buffer or a readable stream.
Also, you can handle rejected fetch requests through checking err.type and err.code.
Only support res.text(), res.json(), res.buffer() at the moment, until there are good use-cases for blob/arrayBuffer.
There is currently no built-in caching, as server-side caching varies by use-cases.
Current implementation lacks server-side cookie store, you will need to extract Set-Cookie headers manually.
If you are using res.clone() and writing an isomorphic app, note that stream on Node.js have a smaller internal buffer size (16Kb, aka highWaterMark) from client-side browsers (>1Mb, not consistent across browsers).
ES6 features such as headers.entries() are missing at the moment, but you can use headers.raw() to retrieve the raw headers object.