I'm saying flush handles aren't about drag, not passing judgement on whether those other factors are bad.

Drag is absolutely one of those factors. Yes, it only contributes a small amount to the overall drag profile of the vehicle, but a vehicle is a sum of its parts ultimately.

It's not a meaningful factor in decisionmaking. Manufacturers went on an aerodynamics optimization spree in the 80s after the fuel crisis. Concepts like the Ford Probe actually dropped handles and all other protruding surfaces in favor of things like electrical touch panels. Seriously, go look at the photos. Even the pillars are flush.

The production vehicles designed after these concepts often used flush pull-up handles for aerodynamics. Those handles later disappeared in favor of the more reliable pull-bar handles we're familiar with because improved CFD made it clear how minimal their benefit actually was for the tradeoffs.

Of course, even if we accept that all the mechanical complexity of flush handles is necessary for aerodynamic reasons, it's not the only alternative to pull-bars. Look at the Volvo EX60 for an example. Designing a flush handle is hard. Tesla spent years working on it. It's not something undertaken for negligible aerodynamic benefits.

What tradeoff is there between pull-up and pull-out handles?

They can't take as much force and they're less reliable. Sometime in the 90s-ish a new test came into force that greatly increased the impact they had to take without unlatching and continue working. The pull bars made it easier to meet because they're secured on both sides.

The pull-up latches also caused issues for people with long nails. In some places spiders liked to nest inside them. Places with snow had issues with a sheet of ice forming over the entire panel, an issue that also occurs with modern flush latches.