Selecting the right FETs as the power switch is important, when switching 5A at 20V.

My initial pick was the Alpha & Omega AON7407. It has very good Rds_on but was lacking at V_drain-source with only -20V max and V_gate-source of only 8V. Better candidates turned out to be:

Diodes DMP3013 with V_ds = -30V and V_gs = 25V; Rds_on 9.5mOhm (V_gs=10V)

Rohm RQ3E120    with V_ds = -30V and V_gs = 20V; Rds_on 8.8mOhm (V_gs=10V)

The following scope plots all show:

Ch1: signal GATES

Ch2: signal VBUSIN

Ch3: signal SOURCE

Ch4: signal VBUSOUT

Math: Ch3 - Ch4 = SOURCE - VBUSOUT to measure Vds

The 1st step to 5V is the initial USB power up, the trigger point is the successful power contract negotiation (blue,light blue and purple trace rise to 20V) and~140msec later, the FETs finally turn on(green trace starts raising) and the input voltage drops slightly because of the 4.85A current draw.

Q1, Q2 = AON7407:

Q1, Q2 = AON7407 with D2 = 7.5V Zener Diode:
Q1, Q2 = Diodes DMP3013:
Q1,Q2 = Rohm RQ3E120:
The AON7404 shows a weird pulse on the math channel when USB-VBUS turns on to 20V and the turn on cycle with 15V gate voltage is not really smooth.
The Diodes and Rohm FETs behave identical with proper turn on cycles.

In the end I decided to use the Diodes DMP3013, as it has extra gate protection diodes built in.