This turned out to be longer than I intended. Apologies.
I view Ethereum as a value network, connecting disparate sets of transactional use cases around a set of core services (like Address, asset records, and transaction functions)
To believe that blockchain makes sense for assets which do not require absolute integrity, you'd need to first accept that there are valuable use cases which having an asset management & transaction layer (L1) serves.
If we establish that there are valuable use cases that attract asset management to L1, at a certain point network effects begin to take hold, and the system becomes "top of wallet".
There are parallels in how you manage traditional finances today - Even if you have multiple bank accounts and digital wallets for USD (e.g., a Chase account, Cash App, Venmo, etc.) you're likely to mentally consider one of those your "primary" account. The important one. The main difference in the value network of Ethereum is that the primary account can aggregate the assets that are managed/transacted through L2 solutions. The L2 solutions leverage the core "identity/asset mgmt layer" of L1, but serve use cases that don't justify the cost of development/operation/transaction on the main layer.
To connect this analogy to the original question, let's imagine that your Bank Account offered direct integration with each of the other accounts you manage - Capturing every asset you held, including the 124 gold you still have on your World of Warcraft account from a decade ago. If the 124 gold were to somehow disappear due to a bug/hack/other integrity issue, your bank account would reflect that. But the important stuff would be there.
TL;DR - If commerce generally moves to blockchain systems at significant scale, there will be an acceptable level of failure on L2 systems to support the convenience of aggregating up asset mgmt alongside the important stuff.