Consider batch allocation strategy for RBT

[Chrinkus]

One use case of my red-black tree had over 140k individual allocations. Would this structure be improved if it had an underlying batch allocation structure?

The array's solution to this is exponential growth but an array of node*'s would require reallocations and data movement. Perhaps a linked-list "pool" of progressively larger allocations?

[Chrinkus]

This leads into a larger discussion about containers taking allocator types.