In the worst case (1 tx per channel) LN would actually double the # of tx. I am really looking forward to LN being really tested, but the thing that worries me

In the worst case (1 tx per channel) LN would actually double the # of tx. I am really looking forward to LN being really tested, but the thing that worries me is that the security model relies on txing before a certain block. If we see a large amount of the traffic move from on chain to LN then we could get ourselves into a situation where there is a run on the chain to close channels. If there is a large market maker that attempts to steal from a channel that is being used to route many others we could see a mass closing of channels which could push the fees high enough to make it cost prohibitive. LN isn't a panacea it actually changes the security model.