Most C/C++ compilers have successfully done tail call optimization for years, and have been covering more edge cases over time. But of course those are fragile niche languages, which can’t compete with Kotlin 
The optimizations are all safe, except for some aggressive ones to cheat in benchmarks – not speaking of core dumps, which have other reasons.
It’s merely an optimization, which (apart from allowing far deeper “recursion” and allowing to implement state machines where the transitions are tail function calls – impossible in Kotlin currently without other-function call optimization) is completely transparent semantically. So it has no business to be mentioned in the code!
What in the world would be lost by checking if a function verifiably called last (with no closures remaining open, etc.) could be jumped to instead?