The current description doesn't really make clear what Manifolds are - if I understand correctly, you've added entities sized (3-5-7-9)x1, with as many connections as possible, and with corresponding fluid volume ?

That's correct. 3,5,7,9 was chosen out of simplicity, though I do plan to make it more 'variable' (think Modular Chests), where any desired length manifold can be built by dragging while placing. The manifolds do have volumes but they are not high enough to be considered as replacing storage tanks - only enough volume to optimize/support maximum possible flow through the manifold with multiple connections made.

Nice ! How did you figure out that volume ?

I did some practical experiments that worked out to roughly matching the volume to the total number of pipe connections x single pipe segment volume. Going too small limits maximum throughout, and going too large makes flow changes slower to propagate across the manifold..

Wow, that must have been a lot of work !

It was relatively simple by using a creative mode map that was previously used to accurately measure flow (used to make that flow chart). With that in place, it was just a matter of setting up some extreme piping configs and figuring out where the bottlenecks were. That flow data also appears on https://factoriocheatsheet.com/

Thanks a lot for figuring out the values for 0.17 !
(Any significant changes from 0.16 ?)

These values also apply to 0.18 (so far). 0.15.0 increased pipe capacity and therefore flow rates by 10x (to where they are today), so the current flow rate behavior should be similar all the way back to 0.15.

Except people already reported electric pumps working slower in 0.17 ?
Or does this only affect pumps ?

Yeah, that only changed the flow rate of the pumps themselves, not the rate of falloff through pipes connecting them.

BTW, what about the issues with flow direction and build order ?
Considering those issues, this graph is likely to be only valid for very specific scenarios?

Build order shouldn't matter, but if you have an example that I can reproduce, I will retest.

Never mind, just retested, looks like the build order issue was fixed in 0.17 or 0.18.
Some people are still complaining about it, but it might be memories dating back to 0.16 ?
https://forums.factorio.com/viewtopic.php?p=475899#p475899

Well since you brought this up, I was retesting, and while build order doesn't matter at all, turns out the pump maximum fell to 6000/s due to some regression. Currently in the back and forth game with the devs, trying to convince them that it is a bug, because you definitely used to be able to hit 12000/s with a pump that is rated for 12000/s: https://forums.factorio.com/viewtopic.php?p=476195#p476195

Wait, i'm confused, what were you discussing before then ?

Except people already reported electric pumps working slower in 0.17 ?
Or does this only affect pumps ?

Yeah, that only changed the flow rate of the pumps themselves, not the rate of falloff through pipes connecting them.

https://forums.factorio.com/viewtopic.php?f=58&t=67037

I tested this [in 0.17.7] based on a blueprint of a pump-bus nuclear reactor design from 0.16 (which used to do the full 12k water/s). The sum of the steam output after reaching equilibrium is around 6.8k per second going by the sum of the steam consumption from the turbines.

The upper bound seems to be 7133 units/s. It's not consistent

I can only imagine 7133 came from an incorrect or otherwise transient measurement. Continuous flow currently works out to dead-on 6000/s. I might spin off a separate mod and/or incorporate into manifolds the fluid box height change to restore 12000/s, but I'm hoping the devs will just fix the value in the base game.

Ok, updated to restore 12,000/s, so now, in response to your question:
"Any significant changes from 0.16 ?"
I can now state 'no' - no significant changes from 0.16 :)

:))

...and now Bilka fixed for next version as well :)