I'm probably mistaken about this, but looking at R9, the phasing connections of the O/P transformer and the grid-anode phase relationships of the valves, isn't that feedback via R9 in-phase with any signal appearing on the grid of the first triode - thus increasing the chance of instability?
Al.