Having the sources is often useful as it helps you to * partition the black box * observe the internal state * observe critical control flows * maybe not 100% diagnose and fix but mitigate an issue
Generally engines and frameworks are bigger than needed. They may be well tested but the sheer volume means there are a lot of corner case defects of holes left. Frameworks are particularly risky as they are supposed to be adapted to new situations and as a rule writing software for future requirements is fraught with problems.
Special purpose frameworks often beat general frameworks hands down in size, performance and cost. Going with a large established framework from a "commercial" vendor is often seen as a safe choice by management and as the article shows it is not. Then on the other hand there have been plenty of teams going down naively the road of custom engines and finding that they did not have to re-invent the wheel but build an advanced gasoline engine and were failing at that. As always understanding your requirements, options and capabilities priceless.