I'll try to give some views on this, I hope it helps bringing some closure to your question.
You seem to relate consensus to "theoretical prices". I think this is a bit misleading. I view consensus as nothing more than the average view across the street for market factors, e.g. the correlation between the Korean KOSPI and the Spanish IBEX, the volatility of Kazakh inflation, etc. So it is really an aggregation of views (i.e. the view of JP's trading desk, Citi's, etc.) for a specific market factor. When push comes to shove, what matters is the actual price at which somebody does trade. Consensus for certain market factors might be meaningless because there hasn't been a single transaction in weeks or months.
I don't really see the discrepancy between the consensus and the desk's mark as "proper" PnL. From experience, there are usually 2 parameters to take into account when assessing discrepancies between the desk and Finance:
- The gap between the desk's PnL and the PnL implied by the consensus;
- The strength of the consensus.
You use the words "consistency", "should not really", "correctly" or "cleaner". Generally speaking, when there are significant discrepancies (what is sometimes termed "variances") between Trading and Finance, that means the market factor itself is thinly traded, thus the strength of the consensus (the evidence of mismarking) is weak. For example, the consensus for KOSPI-IBEX correlation is 70% but your desk is marking at 20%; you go to them and tell them "you're way off the street, you need to remark", but then they say "there hasn't been a single transaction in months, that consensus is not reliable".
The bottom line is, quite often when there are significant variances, Finance has not strong evidence to mandate Trading to re-mark, unless maybe e.g. the gap is really wide, or it has been persistent, or the desk has not re-marked market factors for a long time, etc. It's a grey zone in which nobody really has much evidence to force things one way or another, and thus the implied PnL difference is rather "abstract".
From an accounting perspective, I am not an expert but how can you deal with changes in PnL due to remarking? Somehow you need to balance things out, either by directly reassessing the PnL or recording/releasing some balance sheet reserve. Different choices might affect the balance sheet and income statement in different ways, but it will ultimately find its way towards the bottom line.
Edit: remarking or taking a reserve will both affect the bottom line as well as the desk's PnL, however remarking consists on updating the parameter values and hence has an impact on risk metrics such as VaR.