A way out of the paradoxes here is to recognise a distinction between the crisp and the vague. So the future may crisply exist (being crisply determined by its past events). Or it may only be vaguely determined by what has gone before. So, much like our vision towards the horizon, near term, everything seems crisply predictable. Further away, it all becomes blurred and hazy. And far enough away, it is pretty much undetermined.
Prigogine's book, The End of Certainty, is a good starting reference for this line of argument.
The other key ingredient in the time story I believe is Cramer's transactional interpretation of QM - or simply just taking quantum eraser and other such cases seriously.
This would argue that the future does exist in a quite concrete sense - for micro-events. The other end of an interaction has to be established for an event to "coalesce" into classical existence. So for your eye to absorb a photon from a distant star, this involves a nonlocal (or rather global I would say) interaction across both space and time.
This crisp micro-events level story actually fits neatly into the global "it all goes vague with distance" view of the overall flow of time. It is all about scale.
On the smallest scale, there is really "no time" as the start and end of an energy exchange "happen together" so far as we are concerned from our middleground, classical, observational scale. After all, photons travel at the speed of light and their experience of time would be "frozen".
Then on the largest scale, a global one-way or irrevesible flow of time is being created. This is not a dimension as classically imagined (crisply linear to infinity) but instead this kind of moving cloud of certaintly/uncertainty. Close to the "now" things look fairly crisp and densely woven (either looking back to the past or forward to the future). Then further away, again in both directions, the crisp determinism gives way to vague uncertainty (and hence also, creative possibility).
It is a kind of event horizon effect, but smudged rather than a sharp cut off. From (vague) memory, Prigogine did calculate the drop-off in visibility rate. I suspect it was powerlaw rather than exponential.