Well, there's more than one way.
- You can perform a measurement, though I suppose this doesn't comply
with your definition of 'without code' (since you actually have to
perform a command)?
- Since quantum states are fragile and (at the moment) still not error-corrected, you can just wait for the state to decohere/dephase... if you look at the IBM's device summary information here, you can see that those times (T1, T2) are on the order of the tenth's of microseconds... you you wait a couple hundred microseconds, it's pretty likely that they are unentangled now (though of course, they probably assumed some random quantum state and nothing resembling what you initially programmed.)
Whatever restarting a quantum computer may mean, it probably takes more time/steps than just waiting less than a milisecond.
The entanglement was created with a quantum operation. If you wish to undo and return to the previous state this you have to perform again a quantum operation, which involve hardware actions, which are directed by code. Or you just let the quantum states degrade/get destroyed with non-coherent effects, like noise, etc.
Not sure where your question comes from though, and why you'd want to do something without code, since even the restarting operations probably has some commands being executed. Why would you want to operate a computer without code?