The floating exchange rate between pairs create this risk (i.e. - XBT/ LTC, XBT/EUR, XBT/USD , XBT/Eth, LTC/USD, LTC/EUR,ETH/USD ,ETH/EUR ,EUR/ USD) By this virtue total of 5 simple triangular pairs and multiple complex pairs possible. The root of the problem and goes deeper into system dynamics and is a form of systemic risk. Lets say there are 3 exchanges (X1 , X2, X3). X1 & X2 Deals in all the pairs and hence becomes the major provider for market. One the other hand X3 deals in pairs associated to LTC, BTC, USD and EUR & X4 deals in ETH, BTC, USD & EUR. By virtue of triangular arbitrage between pairs the liquidity of Fiat will move towards exchanges trading ETH. There by leaving X4 with LTC pile up and demand for liquidity as bots would tend to move USD & EUR out over smaller buckets. [Note there is no internal settlement mechanism within cyrpo exchanges (i.e. - X1, X2,X3 & X4 do not talk to each other and cannot come to each others aid)]. Other factors affecting are Speed of movement and cost of movement of Cryptocurrencies & Fiat. By this virtue what happens is X4 is incumbent and X1 & X2 lead the race and X3 is follower [Can be incumbant if BTC gains strength]. There by creating natural monopolies.

Yes point needs some system dynamic and system modeling to be done.