Is it simply that time based effects are always rendered virtually in parallel?
No.
One of the things that might be a stumbling point is that the sound "keeps going" through the blocks. So, the thing that the reverb does essentially starts at the same time the thing that the delay does starts if one is after the other on the grid.
They're not processed in parallel unless you have them set up to be parallel. But, it's also not exactly true that a block has to wait for the one before it to "finish" either. You can simplify the whole thing and think of your entire grid reacting to the input all at once, even though each one is reacting to the output(s) of the block(s) that send a signal into it instead of reacting directly to the input of the Fractal itself.
Technically, some data has to move around, which takes time....but that's not actually specific to digital modelers. Information can't really move instantaneously in a way that's meaningful to audio (and is sort-of an ongoing discussion in physics...a Nobel prize was just given out a few weeks ago for something at least vaguely related...it's also one of the things that Einstein was wrong about because he couldn't wrap his head around "spooky action at a distance"). And while that does technically play a role in audio (and everything else), it actually isn't worth discussing.
If you want to get super-specific, I'm also not sure exactly how Fractal manages the need to move data around between steps. It is possible that it slightly screws up the timing. But, my suspicion is that they do it well. I can't imagine that potential flaws in the exact delay time (which are necessarily going to be less than the total latency of the Fractal at worst) will cause an audible change based on the order of blocks. And, I'm reasonably certain that this potential source of error is handled better inside an all-digital fractal than it would be with 2 digital delay pedals in series with analog interconnects due to the need for more steps of conversion (which each take a non-zero amount of time).
I glossed over all of that in my explanation because it really doesn't matter.