In the real word, more than 90% of the systems do not satisfy criteria 2,3, 4 and 5; almost all of them (exception are systems working in temperature cycles) will never be subject of 3000 cycles (or more) due to a temperature range between minimum design metal temperature and maximum design metal temperature, as they are defined in Projects.
For the rest of (let's say) 10% systems, accurate flexibility is important (and an improvement in calculation was/is necessary) but fatigue under cycles of temperature MDMT-MDT should be confirmed case by case as a realistic scenario.

In the end, how many systems are really subject to such fatigue MDMT-MDT, case for which the benefits of B31J SIFs are obvious? And why we apply them to all systems?

Not the accuracy of B31J SIFs is questionable but the extent of such fatigue for all systems- this is my opinion (and as I've written previously Markl had the same remark before I was born...).