Actual measures show the result is very bad

Modern systems feature a lot of tricks to be faster
The average case is greatly improved
But the worst case is greatly worsened

The problem in this case is most likely cache memory
Data access within the same cache line is almost free 
                        but access to a different cache line is awfully expensive
Here, the two allocations will often fall on different cache lines
Every access to data requires two RAM accesses (with two cache miss)

The "generic" code shown is 20% slower than simple lists
The measure is on the whole program, including I/O