The fact of the matter is unless your company is very mature, getting 3 big features to prod in a month even if that means bringing down some system for awhile is generally much more valuable than getting a single feature out with zero downtime.

As the company and number of devs grow, the probability of a dev doing something stupid increases substantially, so those early mistakes can actually be quite valuable in showing where your infrastructure has gaps.

Let the imperfect code happen early, it often leads to free chaos-testing and more stable infrastructure.