This is by far the most interesting part. I was always taught that to get the interference you had to have identical particles -- and I was surprised when I heard this is how we know to treat e.g. two different photons the same, because two different photons in the same location etc. can interfere with each others' probability. But if there is an extremely clear difference between particles that can still generate these patterns, it means things like hidden variable theories have a much better ring to them.
I've only skimmed the arXiv link briefly, but the method does not rely on interference between photons of two different frequencies. In optical experiments, entangled photons are frequently generated through a nonlinear process in a crystal. High energy photons go in and with some probability lead to the production of two lower energy photons. Energy and momentum conservation lead to constraints on what these can be, such that if you've measured one photon you know all about the other. This is how they are entangled.
In the setup of this experiment (see Fig. 1 of the arXiv link), there are two conversion crystals labeled NL1 and NL2. These convert the pump laser into two longer wavelengths, referred to as the signal wavelength and idler wavelength. The object is probed at the idler wavelength, but the interference depends only on photons at the signal wavelength.
