I'm not 100% sure but I would.make an educated guess that that cluster in the first image for example is a sample of scalable clusters, so throwing more hardwar

I'm not 100% sure but I would.make an educated guess that that cluster in the first image for example is a sample of scalable clusters, so throwing more hardware at it could bring improvements but sooner or later the cost to improvements will call for an optimization or rewrite as you call it, so a bit of both usually. It seems a bit of a balancing act really!