According to a friend of mine who used to be the sysop for one of the world's major electronic music facilities, doing audio in Unix has always been a very mixed blessing.
The good news was that you weren't locked into the overhead of the Mac's graphical interface, Unix boxes generally had a lot more raw processing power than Macs (which is no longer true) and programming was a lot easier.
The bad news was that squashing bugs was so hard that he generally needed to port code over to a Mac in order to find the bugs. In addition, when you require a graphical interface, you can't access processing power dynamically on demand in a true multitasking, memory-protected, virtual memory environment. Yes, the background tasks never slow down or crash the rest of the machine but foreground tasks can never grab more processing power either so all things being equal, a lot of stuff is likely to slow down from what we are used to.
The reason Unix is more reliable is because most applications involve decades of government-financed code that is well understood, has been debugged for years and is available to anybody in the public domain. Unfortunately audio does not benefit from that solid legacy code base.
I would strongly advise people to not assume that audio applications for system X are going be as reliable as what we have become used to on the Mac over the past five years.