if you measure it by percentages then the sub-USD$1 equivalent transaction is already too expensive to be economical, but there is a point in the supply to demand ratio of a good where the cost is virtually zero, of course this depends on the cost of production being low enough versus the volume produced

i'd argue that one of the problems we have with lightning relates to the lack of any kind of market in fee rates

this is also a problem related to the opacity of channel balance as well, and both are based on dumb assumptions (that public relays should have private balances in their channels, and that only actual businesses are going to run LN routing - which is patently stupid because efficiencies are inherently better with specialization)

haha... yeah, it's so early it's retarded... zaps are literally the first almost feasible example of lightning's efficiencies having a visible advantage