Reached through UPDATES. The socket a valve of its kind (its graphic less 16, 96-99) is put on to activate the chamber; the valve does the work (LOOSE_VALVE). This only brings the sparkle back: when nothing lies in the room from the places and the record after the socket is empty -- as when the sparkle went because a valve lay here, and the valve has been picked up -- it puts the sparkle there again (graphic 108 plus the socket's kind).

The record keeps the sparkle's old position, which is what makes it appear in the right place: the projection it then asks for (CALC_PIXEL_XY_IY) is of the record at IY, which here is whatever an earlier routine left there -- the legs' record, in the simulator -- rather than the new sparkle's. The build's sessions never ran this; a scene staged in room $0C, with a valve in the room and then taken away, did (measured): the sparkle came back over the socket.