Why are JIT compilers time constrained? If you have a banking system that runs all day every day, why does it matter if it takes 5 minutes to compile the hot spots at the top tier of optimisation?
Imagine you happen to walk up to your ATM when it's Mother Banking System is in the middle hotspot compile followed by a GC pause. What would you do if your ATM took 5 minutes+ to respond to your Withdrawal Request?
You'd probably freak out and worry about getting ATM card back.
The program doesn't have to stop to run the JIT - it can run it on a background thread.
I would hope an ATM isn't built to use a stop the world GC along with a stop the world jit compiler. I hope you realize how absurd you sound making an argument against jit with a real time device.
Hotspot compilations are blocking? That would surprise and horrify me.
If you have a banking system that runs all day every day, why are you not compiling ahead of time and running super optimization then? JIT is most useful where the application or the target is highly dynamic and you aren't able to compile ahead of time.
If the workload characteristic changes over time you may want to recompile with different profiling data.