(A bit of research later:) It's actually a bit of a mixed bag. The "Operating System Reference Manual for the Lisa" [0] reads on pp. 1-3/1-4:
> Several processes can exist at one time, and they appear to run simultaneously because the CPU is multiplexed among them. The scheduler decides what process should use the CPU at any one time. It uses a generally non-preemptive scheduling algorithm. This means that a process wlll not lose the CPU unless it blocks. (…)
> A process can lose the CPU when one of the following happens:
> • The process calls an Operating System procedure or function.
> • The process references one of its code segments that is not currently in memory.
> If neither of these occur, the process will not lose the CPU.
In other words, non-preemptive, unless the OS becomes the foreground process, in which case it may block the active process in favor of another one currently in ready or blocked state.
