When you are measuring the AC current be careful.
I don't remember how I took the AC readings; I made those observations two months ago. But I really only wish to show that the IV-18 works fine with DC, so the AC measurements are not important.
Although it may be true, I cannot think of why heating the filament with the same amount of AC or DC current would make a difference in display brightness.
With DC, there is a voltage gradient across the filament because one end of the filament is positive and the other is negative. But the grid/segments, when powered, have a similar, large positive charge.
On the negative end of the filament, the difference in charge between the filament and the powered grid/segments is relatively large: more electrons travel from the filament to phosphors on nearby segments. On the positive end of the filament, the difference in charge between the filament and powered grid/segments is smaller: fewer electrons travel from the filament to nearby segments.
To demonstrate this effect, grab an IV-18, twist all the grid/segment wires together, and connect them to +12 volts. Apply +5v to one end of the filament and ground the other. There will be a noticeable brightness gradient across the tube. The brightest end will be the negative end.
Next, reverse the direction of current across the filament. The brightness gradient will be reversed.
Increasing the voltage applied to the grid/segments will reduce this effect and make the overall display brighter. I believe this was the reasoning behind the following entry in the Adafruit FAQ.
At low brightness, the seconds are dimmer than hours. Unfortunately this is an effect of the way we drive the tube. It is not apparent at medium and high brightness
If the dim digit problem were due to driving the tube with DC, increasing brightness would make the voltage gradient across the filament smaller relative to the grid/segment voltage. Brightness would become more consistent across the tube. But in my experience, increasing brightness does not seem to fix the dim digit issue.
Quite the contrary, what eliminates the Ice Tube Clock's dim digit problem is replacing R3 with a jumper to increase current across the filament. That also increases the voltage gradient across the filament, which should (and probably does) increase the overall brightness gradient. (The brightness gradient is still insignificant.) Therefore, the dim digit problem cannot be a consequence of the filament voltage gradient. It must be the result of insufficient current across the filament.
Insufficient current across the filament would result in insufficient heating, and the filament must be heated to emit electrons. The filament is in a vacuum, so it will lose heat through the metal attachments holding it in place. The rightmost attachment is larger; the filament should be coolest near the rightmost digits. The leftmost attachment is smaller, but should still cool the filament. So if the problem is insufficient heating, the problem will affect the rightmost digit most. That's the exact problem people are reporting.
There is a reason behind the claim that VFDs must be driven by AC. Modern VFDs generally have much lower grid/segment voltages, so the voltage gradient across the filament is large relative to the grid/segment voltages. For that reason, modern displays exhibit a much more extreme brightness gradient when using DC to drive the filament. So if someone only had experience with modern VFD technology, the conclusion that VFDs must be driven with AC to produce acceptable results is probably quite reasonable.
It's worth mentioning that both sources cited by neutron spin--the LM9022 data sheet and the Noritake guide--are geared toward modern VFDs which is why they both stress the importance of using AC to drive the filament.
But for the IV-18, as driven by the Ice Tube Clock, the brightness gradient caused by DC is insignificant. Driving the filament with AC would eliminate the brightness gradient, but it's is barely noticeable as is. I have to look quite carefully to convince myself that I can see any gradient at all. From my perspective, there is no compelling reason to drive the filament with AC instead of DC.