I've bought the scintillation crystal, however not the NaI(Tl) one as planned but LYSO (Lutetium Yttrium Orthosilicate). The thing is, NaI(Tl) with shipment is 180 euros, and the LYSO (8 pieces) is 124 euros with shipment, so it's much cheaper. This has some implications:

• The LYSO I've bought is not a single crystal but 8 of them, so I will need to use some optical gel to optically connect them.
• The crystal is naturally radioactive so it messes a bit with readings. On the other hand, it can be said that this natural peak is like a self-calibration (that can be done later in the software). So it's both positive and negative.
• The crystal is not hygroscopic and doesn't need a fabric envelope, so it's a small advantage.
• The peak emission wavelength of the NaI(Tl) and LYSO is the same and is matched with the SiPM, so it's ok.

I've designed a holder for the LYSO crystal using OpenSCAD, visible below. It's the bottom side, and the top will be attached to the PCB. The rectangle in the center is the SiPM array, and the grey thing around it is the crystal. I've also added space around it for the reflective material. The holes are to attach inserts and screw it to the PCB.

The thing is that to maximize light absorption by the SiPM, the crystal should be covered (except the SiPM area because that would block the light) with reflective material. I've added space in the holder for that. At first, I was thinking that the best thing would be some kind of mirror or polished metal. That's what I think upon hearing "reflection" :)

Actually not. To avoid bouncing in circles, it's better to use something porous as a reflective material, ideally PTFE (which is just a sheet of Teflon). It's white. This is the one I plan to buy (1mm thickness).

I still haven't found a shop that sells optical gel. Surprisingly, it seems like it's not available on AliExpress.