Some remarks on the stability problem for linear space automata and semicontinuity of cut point languages