They are array encoded, but you still will be jumping all over the array to do your operations. Putting multiple “nodes” at the same location will mean fewer jumps, and hence fewer cache misses.
1
2 or 3 (let's assume 2)
4 or 5 (let's assume 5)
10 or 11
There won't be any further improvement from using a B-tree, which only scatters the memory further. (if anything, you might consider using a higher-base Eytzinger tree for SIMD reasons, but I've never actually seen anyone do this)