For number 3, do we choose where to place the crash? For instance:

Start Transaction
<START, T>
Write(X, v_x')
Flush(X);
<UPDATE, T, X, v_x, v_x'>
Write(Y, v_y’)

---> Crash

Flush(Y)
<UPDATE, T, Y, v_y, v_y'>
Commit;
<COMMIT, T>
Flush(log);

And determine whether the DBMS can recover using Undo-Only Recovery?