I have never seen a design for a trap diople or similar dual band antenna
for 20M and 30M although my modelling shows it should just about work
although the end section after the trap is very short. Commercial antennas
are all either the normal bands or the WARC bands but not both with the
exception of lossy resistively loaded wide band antennas or rhombics. Has
anyone tried this combination?

It sounds worth trying.

With realistic estimates for the impedance and Q of the trap, what
does your model show about the effective 20-meter SWR bandwidth of the
antenna, compared to an ordinary 20-meter dipole?

Another option to consider might be a 30-meter dipole, with some
single-wire stubs dropped down from it partway along to provide a
20-meter resonance.