Most projects do testing before making a release of course, and for lots of software releases are very reliable. The kernel is an example of this, where Arch packages the most recent "stable" kernel rather than sticking to an LTS version. There is nothing "unstable" about this, in the sense of "unstable" meaning unreliable or buggy.
The key thing is that "stable" and "unstable" have practically nothing to do with the quality or reliability of the distro. Unstable only means that the packages may update in non-backwards compatible ways, so you need to be prepared to make changes to accommodate that. Normally it's something like updating a configuration file, or passing different flags to a command line program, or something along those lines. Maybe you run some sort of service that links to shared libraries on the system, you need to be prepared to relink to updated versions or not depend on system-wide versions at all.
So long as you make sure to keep up with changes, rolling release distros are exactly as reliable as upstream releases.
The other thing to note is that tracking releases may actually end up being more reliable, because you are constantly getting bug fixes and improvements!