It's a very unconvincing tale when the origin story of this service is that the author connected to the wrong backend because he used the wrong port. I can see nothing simpler and more straightforward than numbered ports. Even the SRV record seems unnecesary complex for most use cases.
True! But, you wouldn't just need SRV support in browsers--you'd need it in every part of your dev environment / stack that touches the network. If your backend talks to a database, then your backend needs to support SRV. So you need SRV support in your standard library and your browser.
But at a larger level this tool just makes everything work so smoothly. If you have two services with the same name, it pops up a notification in the tray. If you start a new process with a different name, you don't have to configure SRV; the DNS just becomes available automatically. HTTPS is auto configured. That sort of thing.
I think somebody could one day make something like this that was based on SRV, but it would still need all those niceties to make it work well for developers.