Regarding the replication, there are two main "views". Local replication (cashflow) and Dynamic Replication (greeks).
Local Replication: Asume a -mostly- linear relationship between the ETF and a class of assets. Commonly this is a subset or cluster of the main holdings of the ETF. We are basically replicating the price and therefore the returns. To asume a relationship between an ETF and some other stock is a strong hipothesis but plausible when we actually know what the ETF investments are (so the relationship can hold over time).
Dynamic Replication: what we want to replicate is the sensibility of the asset to some other factors (and the price of course). This means we are trying to create a synthetic asset with the same delta, gamma, theta, etc. so we get the same exposure to the market.
If we can replicate the ETF, we can hedge the position by taking the inverse order on the market.
If we can replicate, we can price the ETF based on the replication price. This is the basic idea behind derivatives valuation -> get a replication portfolio, and if the portfolio gives the same payoff, the derivate should have the same price based on the non Arbitrage Principle.