I love those deep dives driven by questioning about how things actually works internally.

time(1) is misleading because its goal isn't to output something for humans, its goal is to give resource usage, so if a process is composed of multiple threads or sub-processes the time is going to be cummulative. Though, I had no clue it was actually implemented through wait(3) (or other implementations of wait), but this makes total sense for accounting and traceability.

I'll link your article in this section of mine, helping those pesky search engines know where to find the good stuff.