These are, of course, personal views. So far as I can trace, there is zero evidence that financial economics (“FE”) is useful for long-term entities. If there is any such evidence, please tell me about it. Rather, I suggest that the converse applies.
The theoretical academic evidence relates to perfect markets but real markets are not efficient (Shiller & others) and perfect information cannot be available to the whole market (Grossman & Stiglitz, 1980). Further, market prices have no predictive return power (Fama, 1965).
The “FE actuaries” normally deny that there is an equity risk premium, in that any extra return is fully counterbalanced by risk. Well, over periods of 15 years, that is not what is indicated by either actual history or likelihood..
Volatility is treated as risk - contributing to deficits. However, the concept (“today forever”) assumes no volatility, an apparent inherently illogical contradiction. That leads to biased estimators.
FE is mainly concerned with the analysis of assets, with liabilities tending only to be considered as negative assets as rigid rotations. However, real world liability cashflows have their own properties.
Should we use an abstract “risk-free” discount rate? Apart from being empirically unobservable, it wouldn't make cash-flows risk-free (which would be alchemy). Long-term approaches need not be the same as for FE because FE is short-term.
A further problem is that FE ignores “path dependence”; the initial conditions may still be partially relevant.
Using low discount rates just leads to high capital values. All government bonds are risky, the extent being just a question of degree. Major defaults have been seen in the recent past (Greece). Although Japan was thought in 2016 to be considering issuing perpetual zero-coupon bonds, leading to credit risk plus liquidity risk, that was ruled out in a speech by Takehiro Sato on 01 March 2017.
In 2016, it was estimated that 80% of UK index-linked gilts (“ILGs”) were held by UK pension schemes. The position may still be similar (DMO don't possess the data). The real yield fell from plus 2.3% (1995) to minus 1.9% in 2016/17. This was never previously anticipated, being economically irrational, driving circular LDI strategies.
Risk quantification is very poorly captured by scalars, with stochastic processes offering a much better approach. Specifically, stakeholders can and should be given much richer information, permitting better strategies to be followed. It seems to me that accountants and actuaries are using mark-to-market despite the evidence rather than because of it.