It was likely kept longer than needed/useful because they didn't want to run into it. If the rocket is still accellerating it could theoretically catch up to the ejected tower. The probably kept it until they knew its velocity after release would be greater than the capsule's orbital velocity.

We don't need to guess about these things. They keep it until after second stage separation because they might need to actually use it for an abort up through that point. Here's a link confirming that the Saturn V launch abort system was retained until 30 seconds after the second stage separation for exactly that reason: https://www.hq.nasa.gov/alsj/CSM15_Launch_Escape_Subsystem_p... I couldn't find such detailed information on SLS, but it's a similar rocket with the same design philosophy and this logic likely applies equally.

The acceleration of the launch abort system on its own, when it's not dragging the entire capsule with it, is double digit gees. The main rocket is not going to catch up to it and run into it.

Isn't the escape tower designed to have a slight radial thrust component? The problem with avoiding an oncoming booster is the same when triggering the escape sequence for human passengers. I mean, there is an automatic self-destruct, but that must not be the only contingency.

Yes but without the mass of the capsule that 'slight' radial thrust isn't so slight And the CG is totally different. The worry is that an ejected tower might spin/spiral or even take a pinwheel motion rather than fly properly out of the way.

Without the weight of the capsule the acceleration of the escape tower is also much higher. So with even a bit of radial acceleration it should get away from a possible collision course pretty quickly.

It "should". That's not good enough.

Well, the whole point of the escape tower is to accelerate the capsule away from the rest of the vehicle. So the tower accelerating by itself should have no issue getting out of the way.

I don't know how SLS's escape tower works and I expect your right with respect to it. However that's not true of all escape towers. The escape tower used for the Mercury-Redstone missions at least was meant to only work when the rocket below it was no longer accelerating.

This became a factor during the (unmanned) Mercury-Restone 1 incident, where the launch tower flew off when the rocket was still on the launchpad, leaving the capsule behind on the rocket because the capsule was only designed to detach when the rocket was in freefall (it wasn't in freefall, because it was still on the launchpad.)

The Apollo escape tower thrusts at an angle specifically to prevent that.

The lateral componant was to pull the capsule out of the flightpath of the rocket. The total accleration of the tower+capsule was miniscule compared to the rocket overall. If it didn't move left/right a bit then the rocket would easily catch up.

The thinking at the time was that, without the capsule attached, the tower might not fly as planned. It could be unstable in the thin up atmosphere. It might spin. That lateral thrust, withough the mass of the capsule, might cause it to spiral, leaving it floating out in front of the accelerating rocket.

I think they're retaining it until just after the final stage separation, i.e. it will be used if that stage separation goes poorly.