Stefan,
Thanks. Yes I recall these now (from long ago). They are 'quantum eraser' type set ups as I suggested. The findings of ALL these are fully as predicted by Classic QM, simply identifying wrong interpretation and analysis resulting in the flawed conclusions you described. No 'backward causality' or 'non-local communication' is then needed.
To simplify, just run through each experiment carefully and armed with the classical mechanism (rather than using confused statements like; "..projects the state of photon 2 into a momentum eigenstate"!).
So here it is; BOTH 'photons' have BOTH momenta and states, and parts go BOTH (secondary) 'paths' (I can explain that more). If one is rotated, reversed or delayed, then when they're 'recombined' (on a board OR statistically!) there will be CONSTRUCTIVE INTERFERENCE so an interference pattern. Don't forget that just one extra reflection can do this by leaves a state INVERTED.
Now look at say the PR/Nature 2005 review article. Fig 4. As the filter (NDF) is gradually introduced it rotates the state so only DESTRUCTIVE interference results. Though things have moved on in 12yrs there were some interesting comments in the links, i.e; "..many phenomena thought to be due to the quantum nature of light can actually be explained by using a classical electromagnetic field and by assuming that only the processes of absorption and emission are quantized." But then poor interpretation of other findings confounds! The links were then very interesting, thanks.
Also in there, after designating Bosons unexplained 'spin 1'; "Fermions would behave differently because their quantum state is antisymmetric, as reflected by a negative sign in their initial state. In this case the two amplitudes introduced above interfere constructively and the two particles are always found in separate outputs. Interestingly, this 'fermionic' behaviour can also be observed for two photons if the photons are prepared in an antisymmetric state with respect to their spin." (another way of putting what I explained, that they CAN for most purposes be considered in the same way).
Once you've read Bill McHarris's essay watch the IBM Q Charlie Bennett video (referred by Robert Groess below) for the latest analysis, far better than the confused stuff in these papers and even more consistent with ClassicQM, but still hitting QM's 'brick wall' and admitting little progress with Quantum Computers.
best
Peter