> It's doubly odd because as many (I suppose) I firmly believed that faster pace and smaller changes would lead to global improvements (same goes for ajax web..).
A late response, but a response none the less.
Faster response time and changes can result in global improvements, but they don't ensure it unless they consider global implications. See also: normalization of deviance [0]. It's very easy to make small changes based only on local considerations. These can, individually or in aggregate, produce a globally worse system.
A non-programming example, but I think illustrative:
A friend works in aircraft maintenance as an engineer. They use an adhesive to apply patches to aircraft versus bolts and other joining mechanisms (it's less harmful, longterm, to the airframe and less disruptive to the operation of the aircraft being lightweight and not intruding into the airflow). When the aircraft come in for maintenance the worker is supposed to use a disposable plastic tool and some chemicals to remove the adhesive, it is a slow tedious process. So, naturally, the worker makes a local change to improve their flow: a metal tool that quickly scrapes off the old adhesive. Job done, they move on.
The decision was made only with local consideration, but the consequence was an extra 2-4 weeks of maintenance on every aircraft this happened to. Why? The metal-on-metal scraping resulted in damage to the aircraft that required repair.
Unless you're making changes with the whole system in mind, the consequences can go either way with regard to the global optimum, while moving towards a local optimum.
[0] https://en.wikipedia.org/wiki/Normalization_of_deviance