For Libor we have the following Convexity adjustment formula for payment delay (under normal model)
$$CA = P(0,T_e,T_p)\rho\sigma_e^L\sigma_p^L\Delta_e^p(T_s-t_0)$$
where
$T_s$ is the period start date
$T_e$ is the period end date
$T_p$ is the payment date (assuming here $T_p$ > $T_e$)
$P(0,T_e,T_P)$ is discount factor from $T_e$ to $T_p$
$rho$ is correlation between 2 forward rates $F(t_0,T_s,T_e)$ and $F(t_0,T_e,T_p)$
$\sigma_e^L$ is vol of $F(t_0,T_s,T_e)$
$\sigma_p^L$ is vol of $F(t_0,T_e,T_p)$
How would this formula change for RFR compounded (SOFR) rates? A naïve way would be to just replace the Libor vols with backward looking RFR vols and change the $T_s$ to $T_e$ something like this
$$CA = P(0,T_e,T_p)\rho\sigma_e^{RFR}\sigma_p^{RFR}\Delta_e^p(T_e-t_0)$$
wondering is somebody did a proper derivation for daily compounded rates like SOFR.