You can use only one camera for all this and simply move it around with a timeline. Rather than placing cameras everywhere, store the transforms you want the camera to travel to (or use your target’s position/orientation). Feed the timeline with that data and have it move the camera.
If you need something more advanced, you could script a spline system that dynamically places points around the corners and/or avoids obstacles. When connected, these points create a naturally flowing (spline) path whose position and orientation you use to slide the camera along.
This has the added benefit of having a huge amount of granular control, including the ability to interrupt the camera movement organically - something that Set View Target with Blend does not provide, since it’s a fire & forget solution that’s not even aware of its own interpolation.