Not really, no: jumping around the timeline wouldn't benefit, and load times per project would be astronomical (*could* be, anyway), reducing usability. It would also complicate code ("is the data in cache or on disk?"), thus increasing the likelihood of bugs.
There's also increased chance of memory use clashes with large, sample-based plug-ins, reducing stability (try loading even a smallish sized BFD2 kit into memory and watch an 8GB machine crash and burn).