So the core idea is to move to arbitrarily wide logs? Seems good in theory, except in practice it just defers the pain to later, like schema on read document d

So the core idea is to move to arbitrarily wide logs? Seems good in theory, except in practice it just defers the pain to later, like schema on read document databases.