The 68000 design is lacking the necessary parts for accessing an address and having the OS recover and handle it when it is pointing to void and nothing or just a part of the address map where nothing lives.
The 68010 added the necessary states and instruction to work together with an external MMU to do this. The other interesting solution was to run _2_ 68000 cpus in lockstep where the hw was designed to detect this and switch to the other cpu to run the offending instruction (there would obviously have to be some cleanup and mapping change before it could do so). Can't remember the name of the system that did it, but it was a Unix-like IIRC. Someone will now find a link to it in 5... 4... 3...
Early Apollo workstations.
The best laid plans of mice and men are held up in the legal department.