There is big support of these models from consulting world. If it doesn't work than it just means we need further micro components to make it work but never try back to integrated system.

Some two bit code monkey screeching "it's not my fault my code shit all over the production database when something upstream of it got a little wonky, I coded to the spec I was given" is no different than some dumbass middle manager or MBA at HQ who's intentionally making the system more shitty for negligible, but measurable, savings so they can fuck off into the sunset with their bonus for hitting KPI before it all comes crashing down.

Ergo, any decision which saves dollars and time in the present but increases future risk looks good... in the present.

It's always surprised me that businesses don't attempt to quantify risk in an effort to produce a more balanced decision function.