Agree about when the engine is running. No issue there. When cranking on starter the RPM will vary as cylinders go into compression but the varying RPM will not match the varying load on the belt as thats dependent on valve springs not compression. Initial starting torque depends on this: https://en.wikipedia.org/wiki/Moment_of_inertia which will be a small value for the timing system.
It is analogous to a bolted or bonded joint, e.g. the belt is a plate and the individual teeth are fasteners. For joints with constant properties, which is what we have with the belt and cog, the end 'fastener' will have substantially higher loads and the reduction over each fastener is not linear.
Forget the compression. It was an example that was suppose to make things a little clearer. Apparently not. But I already stated this, "The torque applied at the drive pulley generates a tension in the belt. That tension generates a torque in the driven pulley. That torque is either sufficient to over come the forces [all forces] resisting rotation and accelerate the effective moving mass of the valve train or not. [ I.e. starting from rest there must be acceleration.] The varying forces in the valve train do not alter the tension in the belt, they alter the acceleration of the cams, valves, etc. " I my hast I posted the last statement which was in error. The varying forces in the valve train will effect belt tension. If the forces opposing rotation decrease then one would expect belt tension to decrease; if they increase, belt tension well increase. I presume you are ok with that? I will insist that total mass plays a roll, but will also concede that if the accelerations are small, the forces associated with acceleration will also be small. Mass will determine how fast the system response to changes in force. And, since the drive pulley is connected to a lot of moving (rotating) mass (crank, pistons, rods, fly wheel,... it will not response very quickly to such changes in belt tension. On the other hand, if the drive pulley and associated mass were very small, belt tension would remain much more constant as the drive system would accelerate or decelerate very quickly in an effort to maintain constant tension. I agree, if you throw that into the mix, it only add another source of as asynchronous speed variation to effect tension.