JACK was developed in part because of the absence of a viable plugin API on Linux. It allowed people to "glue together" whole applications rather than load plugins into a host. This is pretty cool, no question. But the session management aspects of it are not that cool, and despite both the JACK session API and the Non session manager and other things that falktx (Filipe) have done, the situation for users really hasn't ever gotten to the point where reloading a "JACK session" comprised of many individual applications is as easy as it ought to be.

By contrast, we now have a fairly excellent plugin API (LV2) that is supported by several hosts, a number of which offer up quite different models of plugin loading, control, management and usage (Contrast, for example, Ardour, Carla, Qtractor, Non-foo and jalv). I think that most users are better served by developers creating LV2 plugins and the users loading them into one or more hosts of their choice (the so-called "monolithic environment" approach). This allows save+reload to be simple, fast, and reliable.

There will continue to be times when using JACK to glue together applications is still a really useful thing, and I hope that JACK will continue to be there to support those times. But I don't think JACK is generally the right solution for most users, most of the time. The one big positive it still offers: multiple applications can access the audio device at the same time, with low latency and synchronous behaviour. This is an important feature not offered by ALSA or PulseAudio, and for that reason alone, JACK remains important. It also the case that JACK is the only cross-platform system for doing inter-application audio.