One of the problems with BAT is that ERC20 doesn't scale so as they get more users, they will need to cut back on the frequency of payments. Also, a lot of that value is absorbed by the Ethereum network as transaction fees (GAS).
When you combine the complexity of the system, the constant need to try to maintain backwards compatibility and the sheer number of people involved in the project, you have every reason to be skeptical.
This is a 2 year old codebase that has had a paid audit and tons of eyes on it. Just this week, someone discovered and announced an issue. While the fix is relatively trivial, the amazing thing is that it went unnoticed for so long.
I can only imagine ETH2, which is a brand new codebase, whole new consensus model, using cutting edge research (zkSNARKS) and shaped like a rube goldberg machine, will be full of issues nobody has even dared to imagine. I couldn't imagine trusting billions of dollars of value to that literally over night.
Aside from major technical challenges that are specific to Ethereum, a major problem with sharding a blockchain in general is that it somewhat obscures the reality that each shard is in fact a separate blockchain with separate network participants and so it might as well have a separate unit of value and not be constrained to a single token.
The ideal solution IMO (to really promote decentralization) is to have multiple independent blockchains and networks interlinked together via open source decentralized exchanges - This will mean that any token's value could be instantly determined relative to that of any other token in a trustless way by polling decentralized exchanges.