I listened to most of it today and will finish it on my drive to work tomorrow.
A "killer app" like how Ruby has Rails might be needed for Scala to catch on. Lift is the main current web framework in Scala but a web framework probably won't be its killer app because so many other great web frameworks already exist in other languages.
One or two people from the panel were hung up on Dependency Injection (this really bothers me, as DI's usefulness arose from the limitations in the Java language not found in many other JVM languages). It was pointed out that traits can eliminate the need for creating a separate interface just to have a swappable class for testing.
The complicated things done with the language will be done by those creating libraries and the normal developers will merely interact with those libraries.
Scala is a flexible language very suitable for creating DSLs.
At least panel member admits they're still grasping the language itself, even avoiding some existing libraries while trying to understand core concepts... yet another panel members tries to claim that Scala is simpler than Java because most constructs are libraries rather than extensions to the language.
Yes, although Lift seems to be very good, bordering awesome (it uses continuations I think), the "killer app" cannot be Yet-Another-Web-Framework... actors and erlang-style concurrency in the JVM can be really compelling but there's a couple of actor libraries for Java too...
Internal DSLs can be a strong point, backed by Scala "magic" and interacting with Java they will be easy to adopt as only the DSL creator needs to really know Scala. It won't be "the one killer app", but it would be lots of solid apps.
Actually, it got me thinking: how I can make a difference with Scala right now in my work? A low hanging fruit could be XML processing, because is really trivial to write the code... there's also a complex logic than would benefit from an DSL, but I guess it would be harder to convince everybody