"Bjarne Stroustrup bolts everything he's ever heard of onto C to create C++" was validating. A friend who has worked on desktop software in C for 30 years has it figured out IMO: he uses the C++ compiler, but always defines no-op constructors. Instead he defines an init() method on each class, which he calls explicitly. Essentially his only use for the C++ compiler is to sort out inheritance and build a correct function dispatch table.
This gives up the significant benefits of RAII, which are made significantly safer and more useful through the use of references and std::unique_ptr. By using initialization methods rather than constructor methods, you create a situation where your destructors don't necessarily make sense and can't be relied on to be dealing with initialized objects (thus increasing the amount of boilerplate code that can fail) or you don't use destructors at all (at which point you regress to the "boy, I hope I scope-guarded correctly and cleaned up along every path of execution!" state of C).
Writing C++ without using C++'s idioms is usually, if not almost always, creating more technical and business risk for the writer and the consumer than using C++'s idioms in the first place. (That some of these idioms are conceptually very difficult is a very valid criticism of C++, and one that Rust in particular seems well-positioned to be better about, but I'd take the bear traps of C++ over the land mines of C almost any day.)
