Do you mean "thread local passed to another thread by mistake"? There is an option to make data global instead. Unless it is a Unix pipeline kind of design, independent threads can populate their working set (from file/socket/whatever) into thread-local memory.

It depends on whether shared-memory is a design requirement or an implementation artifact (Erlang does just fine with a largely shared-nothing model).

It also depends on whether a program runs for a short time, or for a long time. If you are running for a short time, why not just avoid the GC entirely; "manage" memory manually by malloc-ing and never free-ing (free-ing too takes time, and doesn't make memory available to other processes anyway).

The typical approach here as far as I'm aware is to use a deallocation queue - frees of an object that was allocated on another thread are freed by putting that object in the original thread's deallocation queue. When that thread gets an opportunity, it frees everything in its deallocation queue to reclaim that memory.

Manually, or automatically?

In D the GC heap is shared across registered threads, so it supports values allocated in one thread and passed to another. This particular passage from one thread to the other is guarded by a "shared" type constructor, so as to have a type system guarantee.