Thanks Alex - this is a superb piece of analysis.
Given the number of places L: vars are accessed (model XML, html/js, wasm, simconnect, typescript) I’d suggest this glitch is an unintended consequence of running the sim from the World Map and missing some of the details that might get affected. Maybe Asobo should include the L: var load in the new sim initialisation / plane load sequence.
Would having an L: var purely for the purposes of working out whether the L: vars have been loaded (i.e. L:L_VARS_LOADED=1 in systems.cfg [LocalVars]) be the most robust workaround?