As I mentioned in my initial post, I wanted to review the design with fresher eyes before I send them off for prototype manufacturing. I was pretty confident that I handn't miss anything, but oh boy... there was some major mistakes in that design. Here were the major ones:

1. Needed to use HSS configuration for NCV68261

The ideal diode circuit has been correct to use back-to-back NMOS for HSS + ideal diode configuration:

The initial design actually had the control circuity powered from the charger output, and so initiallly I turned off the 5 V boost converter itself so this HSS mode was not necessary. 

But in the design in my initial post, the control circuity is now powered by the 5 V boost converter, with its bypass mode enabled for power saving. In this case, the additional NMOS is necessary since even if the initial lone NMOS (now Q3) is off, the body diode can still conduct.

2.  Flipped footprint for DMN3016LFDE

The way the PCB is layout, all of the SMD components are mounted at the back. It can get pretty confusing, especially since I'm used to the 'normal' way of them being at

the top. 

Previously, I flipped the gate and the smaller source pad, when it should have been the other way. 

These are the two big errors. There some smaller stuff to fix like wrong part numbers, 5 V boost convert NC pins not being connected to ground, etc. Other than these fixes,, there is one more major change, and that is using a real power mux for the inputs. I decided to use TPS2120.

And that's it. I will wait another few days, redo the checks and see if I missed anything this time...