There's been >2 hyperthreads in other server CPUs before. IBM and Sun/Oracle now or in the recent past have sold systems with 4-8 way threading. Of course, these would be niche systems even without lots of threads per core, so one's definition of a "common" workload would need to be adjusted accordingly. My experience was that they worked really well for big databases and Java app servers, less well for other uses.

"dead time" occurs when a thread is stalled. The most common reason is a cache miss. So if the running applications have sufficient cache misses, additional threads may help - provided that the applications are multithreaded.

When you'll have 4 cores per cpu, there will be usecase fot it. Generally IO is catching up with cpu performance, so it makes sense to do this.