Yes, it's when small amounts (i.e. individual drops) are added slowly. It's called auto-zeroing or something and supposedly it's a "feature, not a bug" as far as I can tell. This is how it works on my SF-400D at least, not sure about on others: In order to prevent erroneous data from i.e. dust settling on the tray between taring and weighing, very small fluctuations from zero will be ignored. Once a weight is registered, small fluctuations are then counted.
In practice, this means that:
If I tare, and then add a single drop, it will not be registered. I can wait a bit, then add another drop, which will also not be registered. Wait a bit...add a drop, nothing. And again, and again. No matter how much I add, as long as it's one drop at a time and I wait a second or two in between for the reading to "settle," it will continue to read zero. Once a weight is registered (if I add a few drops in quick succession before it settles, or anything heavier than about .02-.03g), then it registers just that last amount that was added, but all the previous single drops are just gone. Poof. As if they never happened. But then, after it gets a reading, further individual drops will be counted; auto-zeroing only happens right near the zero point (after you tare).
Interestingly, if I add an object (a nickel or whatever, I usually just use the cap from the bottle I'm working with) so that it reads a weight -- say, 5g, and then I can add a drop and it will read it -- 5.02 on the display, for example -- but then if I remove the nickel, the .02 that had previously registered, will just "disappear." Once the nickel is gone you would expect it to read 0.02, but instead, now that it's back near the zero point, the auto-zeroing feature kicks in and sees that .02 as "erroneous" so just makes it vanish. I can even put the nickel back on and it will read 5.00 again, as if that drop I added -- and the scale registered, initially -- never even happened.