Rio:
You talked about 2x6 t&g roof decking on 4x8 rafters with 2 - 3/8" lag screws through each 2x6 and into each rafter. To get a feel for what your engineer was trying to achieve, try this experiment. Exact material is not real important because we are just trying to get a relative feel for the situation and the point I’m trying to make. Using a 6 or 8' long 2x6 or 2x8, attach one end to a rafter piece, not too near the end so as to eliminate end splitting of the deck piece. Clamp the rafter piece down and pull on the free end of the deck piece, in the plane of the deck, trying to torque the deck piece w.r.t. rafter piece, i.e. changing the 90° angle btwn. the two pieces to 95°. Keep the various dimensions the same so the only change is the fastener size, and apply the pulling force with a spring scale. Use 2 - 16d common nails, 2 - 20d or 30d common nails, 2 - 1/4" and 2 - 3/8" lag screws, lengths should be more than two times the deck thickness. The lever arm btwn. the two fasteners, on this resisting couple is about 3.5" for the 2x6 and about 5" for the 2x8 and only the fastener size changes. The lever arm to the applied force, at the spring scale, would be about 5.5' or 7.5', and doesn’t change. Certainly, you should see the larger fasteners giving larger resistance to the twisting action.
The way this deck structure works as a shear resisting diaphragm is to sum all of these individual connections (resisting couples), which are preventing the decking from racking w.r.t. the rafters, i.e. keeping the whole diaphragm square, or from parallelograming. You can actually force and feel this parallelograming on poorly nailed decks, both on the whole deck at the free edge or at the individual joints deck to joist. There is some friction btwn. the decking pieces, but it is very difficult to put a reliable number on this, given shrinkage and the like. Gluing these connections would be an improvement, but code approval may be questionable without a specific testing program. Also, rim joists and blocking are important to prevent the rafters from rolling over due to this loading, as the system transmits the load to lower level structure.
Sheet goods are very good at transmitting this kind of loading. Rather than doing this through many discrete resisting couples, the sheet goods do this as a continuous panel, and we talk about the loads or stresses, in plane, as a shear field, or as a shear flow in #/sq.in. The experiment here is to stand a 1/4 or 3/8", 4x8 sheet on edge, with its lower corner against a sole plate; and push, in plane, on the upper opposite corner. The sheet will not parallelogram, but it will buckle, out of plane, under sufficient load. Thus, the need for strict edge nailing schedules, and somewhat lesser nailing in the field, to prevent out-of-plane buckling under the shearing load. The discrete condition here is primarily the summing of the edge nails and their spacing. And, you will see tables which allow greater shear loading on a given panel with tighter nail spacing. Also, stud or rafter spacing comes into play in this buckling problem, so 16" o/c spacing will give higher values than 24" o/c spacing for a given sheet thickness; and this system too can be improved by gluing the joints btwn. the sheet and the rafter, joist or stud. In each case you are improving the system performance by transmitting the forces or stresses as more uniformly distributed shear flows rather than at discrete, relatively weak resisting couples.
The ultimate here is the thin skin on an airplane wing, a shear panel btwn. the ribs and stiffeners. Here we control the skin buckling, but it still buckles several magnitudes of the skin thickness, something we can’t tolerate on most of our structures, and the forces and this slight buckling cause the skin to become a diagonal tension field btwn. the ribs and stiffeners, rather than a rigid shear panel. In effect, the ribs are like the top & bot. chords on a floor truss and the skin (diagonal tension field) is the truss diags. And, now the skin is a carbon fiber composite, only mm thick, and the whole system is bonded to provide the connectivity.