You should qualify what you mean by "general classes of problems".
Perhaps you are thinking of the creation of software-development tools, or
general-purpose/'commodity' tools (like word-processors or spreadsheets),
or real-time process-control applications, or multimedia, or network
management applications - situations where you really are dealing with
'objects' and where performance is paramount.
I've designed or built a number of budgeting / accounting and other types
of general business or database applications. I currently support systems
to create and process samples used in calculating the U.S. Consumer Price
Index. In these types of 'business' applications I work with, I have no
need or desire for the 'heaps & stacks' of complexity you're talking about.
The machinery is totally irrelevant to these types of applications. While,
it might be nice to improve performance, the cost of having people who
understand the machinery, or build optimized, special-purpose objects from
scratch *far* outweighs any performance advantage of working at this
level.
My impression is that the primary goal of object-oriented development
environments, at least as far as the business world is concerned, is to
create reusable, general-purpose objects that can be easily extended to
handle the particulars of a new problem to be solved, while hiding as much
complexity as possible. Portability and simplicity of maintainance (mostly
meaning cheap, entry-level programmers wherever possible) is far more
important than performance for these types of applications (which still
account for the bulk of the world's computing, though that is changing ;)
Mark Crosby