I understand the proposition, it just doesn't make any sense. Let's say there is an error in transcription, and you make a faulty transcript that gets deleted b

I understand the proposition, it just doesn't make any sense. Let's say there is an error in transcription, and you make a faulty transcript that gets deleted by NMD. So... what? What consequence does this have for the cell? It just means a transcript was lost; the cell has no memory of this event, once the transcript is recycled and its protein byproducts digested. This is why transcription can be less reliable than replication, because transcripts are disposable and mutations in them have very little consequence. If you mean that it is better to have a fault-tolerant system where potential errors are masked, this advantage exists in spite of a decent rate of error in transcription, not because of it. That is because errors in the germline are fundamentally different from errors in transcription - they are constitutive, present in every cell, and the cell cannot escape them merely by discarding the transcript. Now, presumably mutations that would be horrific germline mutations (a nonsense POLE mutation, let's say) happen all the time in some transcripts in some cells. So what? In what way has the "system" (the organism) been tested for the effect of this mutation at the germline by one shitty transcript in one cell, which was immediately discarded in favor of another, working transcript?