I would debate that, because ST do "offer" the whole thing: USB, FreeRTOS, ETH, LWIP.
They just chose to not integrate them properly. It's in bits around the place. Espressif OTOH did pay somebody to integrate all this and it works out of the box. Not bug free (some subtle bugs in their TinyUSB code) but a lot easier.
I would still use ST for commercial reasons (e.g. likely production life of 20 years on these "industrial" chips, chinese political risk on the key component, etc) but I can see why people go the ESP32 route. They save themselves a year of intense R&D.
BTW, my project started off as a hardware (circuit diagram) copy of an ST devt board (almost everybody does that) but still a vast amount of work was needed to get it to work properly.
RTOS itself is not an issue. It needs software to be written to be thread-safe, but it is basically easy and it is how one should be writing code anyway for embedded work. Variables on the stack, etc. I'd say the use of RTOS is the easiest thing on my whole project, and having an RTOS has made writing all the other stuff vastly easier because you don't have to create complicated "main loop" and state machine structures. You can also avoid a lot of ISRs, or at least keep ISRs very short. To create a new task (thread) you just write it, and set the priority, and the amount of stack, and off it goes. At the end (which should never be reached) you put a loop. 5 minutes' work. RTOS is a really brilliant thing to have; a huge timesaver, and enables easy creation of complex products.
I wonder what the OP is trying to do. If he's trying to build a serial to USB converter, he will for ever be dependent on ST to provide windows driver support for the VCP, whereas for the 2102, FT232, etc chips this is pretty well assured.
