OK I disagree with this pretty fundamentally. It's not a "pass" or "going easy". Here's a THRED on backtesting exceptions ....
The "backtesting exception" is a euphemism for "the VaR model didn't work", as they often don't. You have to record one every time you get (broadly) a movement in the "hypothetical P&L" on any given day which is greater than the 99% 1-day VaR. Whoa hang on ...
"hypothetical P&L" basically means, oversimplifying wildly, "the movement in the daily valuation of the trading book". It differs from "actual P&L" because you're always *trading* that book, so your actual prices achieved might be different from the daily valuations....
"99% 1-day VaR" is basically a standard risk measure; you build a probability distribution of what you what you think the daily P&L could do, and this is the 99th percentile. So you would expect a "backtesting exception" 1 day in 100, or two or three a year.
If you get the expected number of backtesting failures, then that tells you ... well, unfortunately, not very much, because VaR isn't all that great a measure, and in particular the 99.9% loss might be a *LOT* bigger than the 99th percentile. Be that as it may ...
... in any case, if you have a couple of dozen backtesting exceptions, then that is quite bad - your VaR model is not working even in its own terms. For this reason, if you get a lot of failures, the regulators will put a penal "multiple" on your trading book capital requirement
But! I used the word "penal" there because that's what it is. The multiplier regime on backtesting exceptions is not really justifiable on prudential grounds - it's not related to the actual quantum of the mismeasurement of risk. It's meant to be punitive and to give ...
... the bank the incentive to sort the model out, quickly. If the reason for the exceptions is not "VaR model has calculation errors, coding problems or bad data", there are always provisions to get your exception ex'ed out. 
(for example, this happens when the loss comes as a result of a rogue trader or other operational screwup for which you are already being penalised on operational risk capital; it doesn't get double-counted against the VaR model, which there is no need to change)
So what's happened since February is that the industry's VaR models have been hit with a set of volatility shocks of absolutely unusual weirdness and severity. Because the models take time for new data points to effect the distribution, there have been dozens of backtest fails
Does this mean that the industry can ignore the data? A thousand times no! It's data! It's showing that the 99th percentile was in a different place from where it was. The models need to reflect this and potentially to be recalibrated. But on the other hand, is there really ...
... much to be gained by hitting the entire industry with a punitive capital penalty (and thus, not even giving anyone a competitive disadvantage) to incentivise it to do something everyone knows it has to do anyway, sucking up operational capacity that's needed elsewhere and ...
... making trading desks take capital away from lending operations, in a quantum that's explicitly unrelated to the prudential risks? During a period in which most trading operations have actually made large profits?
I can totally see why the regulators have done this. To be honest I think the FINMA approach (simply ex'ing out all exceptions since 1 Feb) is not great public policy and the BoE/ECB approach (using flexibility elsewhere in the capital requirements to offset the VaR multiplier..
... is better. But it's not just letting the banks off the hook; it's recognising something that was always in the regulations - that VaR multipliers for backtesting exceptions are mainly punitive rather than prudential. And as I say, trading books are making money right now.
