Well, as I understand it the problem is if you measure the slit the photon passes through, then the superposition is no longer "left slit or right slit", it's "

Well, as I understand it the problem is if you measure the slit the photon passes through, then the superposition is no longer "left slit or right slit", it's "left slit and sensor or right slit and not sensor", and the slit/sensor paired configuration can't interfere with itself anymore because the sensor state is different. In other words, this is what you'd expect to see collapse or no collapse. For instance, consider https://en.wikipedia.org/wiki/Delayed-choice_quantum_eraser https://en.wikipedia.org/wiki/Delayed-choice_quantum_eraser which demonstrates that the sensor state has to be part of the superposition instead of causing an irreversible collapse: if you store the detected slit until the photon hits the back wall, then delete it, an interference pattern emerges once again.