There’s something qualitative and important that happens when the error on T⟨A⟩ (aka ϵ) is smaller than the amount of time it would take event A to cause anything to happen (e.g. smaller than one network latency): that means that we can be sure that events that are timestamped before T⟨A⟩ cannot have been caused by A. This is a rather magical property.
I’m suspicious of any distributed system design that uses time without talking about the range of errors on the time estimate (i.e. any design that assumes ϵ==0 or even ϵ≥0).
