With even a low to moderate load, the volume of the pressure cell (i.e. fire-formed brass internal case volume) effectively doesn't change, regardless of whether the brass started out virgin, or was already fire-formed. So the volume of the pressure cell between the two types of brass is effectively about the same. What does change is the amount of energy required to expand virgin brass out to fit the chamber, versus the energy required to expand fire-formed and re-sized brass, which starts out much closer to the fired pressure cell dimensions. This is because we typically only push the shoulder back a thousandth or two, whereas it might grow 4 to 5 thousandths (or even more) on the first firing.
In my hands, velocity in small to medium cases such as .223 Rem and .308 Win almost always increases by 10-15 fps or so in fire-formed versus virgin brass. I've always attributed this slight increase in velocity to the fact that the small amount of the total energy generated by powder combustion that goes toward expanding the brass to fit the chamber with a virgin case is free to contribute to projectile movement in fired brass, which is already closer to fire-formed case dimensions (pressure cell). I'd guess the cause of your velocity change has to do with something other than a change in pressure due to the change in effective case volume between virgin and fire-formed brass. If the starting size of the brass actually represented a bona fide pressure cell volume, it wouldn't expand to fit the chamber as readily as it does. Brass can certainly expand further with repeated firings, but the amount it grows after the first firing is only a small fraction of the amount it will generally grow on the first firing with a reasonable load.
There are many things that could explain at least part of the velocity difference you observed, such as a change in temperature or other atmospheric property. I'll acknowledge that 30 fps seems a bit large for IMR 4350, which has reasonable resistance to thermal variation. How many rounds did you fire in your two velocity determinations? Statistical variance could also be a part of it if the group size was small and/or with only a single replicate group for each brass prep. You could spend a lot of time and effort trying to specifically identify the responsible variable(s). However, it's probably simpler to carefully note and record the change, then move on and determine whether the "new" velocity in fire-formed brass (2790 fps) remains constant, and whether you need to tweak the charge weight to re-optimize the load for fire-formed brass.