This change seems a bit heavy handed to me.

By moving rigidbody_update_simulation into the sub step loop, we are now validating the constraints and updating collision shapes of all objects on every substep.
So a lot of things besides the intended effector update is updated on each substep now.
I don't think much of that data changes either, so this leads to a lot of unnecessary updates.

Wouldn't it be better to create a new function with the effector specific logic from rigidbody_update_sim_ob and state that forces not inherent to the rigid body simulation need to be re applied on each sub step?

Yes this was hasty fix, thanks. I'll look into splitting this up.

I was considering a versioning fix that scales down existing forces to compensate for the increased "effective strength" and avoid breaking existing setups. But effectors are used in other simulations too of course where forces were not affected by substeps, so trying to adjust user data like that would probably do more harm than good.