Been doing this for a long while now, only time I ever got ooms was in the editor and also a JT (i think) scenario where you had to run a different scenario first to prevent it (can't remember which one off the top of my head now).
I wasn't aware of this workaround, so trying it now, but having said that, I rarely get OOM in normal gameplay, only when editing.
I thought that shutting the computer down and disconnecting from the power supply would have cleared the Page File? Is that wrong?
So I have finally cleared up the quick drive that was giving me repeated OOM CTD's. Again shared in the spirit of "might be helpful" rather than "this is the one true answer". Rebuilt the quick drive from the bottom up completely new adding one AI consist at a time and then testing - and Yes it was a very slow process that took all weekend. What I discovered were two distinct causes of OOM. The first was where I had asked an AI loco to do something that permanently blocked my path. I had a coal train that emerged from a junction ahead of me and waited in the loop for me to overtake at the station platform. What I hadn't picked up was that the train length was so long that it traversed the points at the entrance to the loop blocking the points I needed to direct me to the platform. The scenario didn't need to start for this happen. The checks that ran as part of the scenario start up picked this blocking up and crashed to an OOM before the scenario even started. Looking in Logmate output showed a number of resolved blocked paths so path blocking of itself is not the issue where it is resolved by the blocking train doing something to get out of the way; like crossing a junction or reaching a portal. The second was where I started the scenario at an intermediate station on the up line. On the down line there are AI consists ahead of me and AI consists behind. As the scenario starts all of the AI consists within range behind me start up AT THE SAME TIME. When building the scenario I had of course built the sequence of consists so that they started in a cascade as I approached and met the start distance. Because the consists all started at the same time they started to run into each other where there were no signals to prevent this - result carnage and a crash to OOM. By making sure each section block only contained one AI consist ( as per real railway practise) problem avoided.
