I have considered
* scanning a linear array of BPW34 photodiodes, in a similar spirit to a scanner to cover a plane, each photodiode going to its own "MCA circuit" (TIA->cheap audio codec like those from Everest Semi). Either direct measurement of generated charge pulses or covering the photodiode with phosphor on aluminum foil or so
* cloud or bubble chamber (cloud chamber is less dense and will generate fewer events, so probably bubble chamber): instead of needing a large 2D or 1D array of parallel circuits, we image and track generated charged particles and use the trajectory starting end (less curved) to determine the source direction!
* consider X-ray crystallography, an incoming straight beam can diffract in many directions on a monocrystal. rotating say a silicon wafer, and measuring the incoming photon energies with one or more photodiode/MCA circuits we can assign a source likelihood distribution by keeping track of the orientation of the monocrystal. akin to sparse sampling but instead of masks its diffraction patterns.
If you have better ideas or variations in mind, let me know!