Priming, and getting the air out before the first measurement
Filling a channel network completely, and why it is harder than it sounds.
Priming is the deliberate step of flushing every last pocket of air out of a manifold's channels before the first real measurement is trusted, and it is harder than it sounds because a fast-moving liquid front can race past a side branch or a sharp corner and seal a pocket of air in behind it, where filling a jug would simply push the air up and out.
Priming a dry hand pump
An old hand-cranked well pump will not draw water once its chamber has gone dry, however vigorously the handle is worked, because its seals need a column of liquid already present to grip and lift, and cranking against air moves air back and forth uselessly. The traditional fix is to pour a cup of water down the spout first, filling the chamber directly before trusting the pump to draw from an empty start. A manifold's priming step asks the same favour of the system: fill the network deliberately before relying on its normal cycle to move anything, since a network full of air can no more measure a liquid than a dry pump can draw water.
How a closed network traps air
An open container is forgiving because air leaves straight up through the same opening the liquid enters. A closed network of narrow channels removes that escape route, so air trapped anywhere inside has to find its own way to a vent or outlet, travelling against or alongside the incoming liquid. If the liquid front reaches a branch or dead-ended side channel before the air there has left, it can seal that pocket shut, and from then on continued flushing through the main channel simply flows around the bubble.
Speed makes this worse. A fast front follows the path of least resistance through the largest channels, arriving at downstream junctions before slower side branches have filled, and once the main channel is full and pressurised, air left in an unfilled branch has nowhere to go. A slower fill gives air time to be carried ahead of the front and out through a vent. That is why priming procedures specify a rate as well as a volume, and why they call for flushing the network with more than one working volume: a single bare pass tends to leave air in exactly those corners, while repeated passes give each trapped pocket another chance to be caught and carried out.
Letting the air expand its way out
Some corners are shielded from the main current however the fill is timed, and a different approach attacks the trapped air directly. Drawing the primed network down to half an atmosphere doubles the volume of any bubble still hiding in it, and a bubble that has grown that much is far more likely to find any path out, however indirect, and rise toward a vent. When the vacuum is released, the surrounding liquid closes back into the space the air has vacated. This does nothing for a pocket with truly no path out, which is a geometry problem for the designer, but for the more common pocket that is merely slow to escape it can finish what a mechanical flush left incomplete.
Priming is the one point in the cycle where every part of the channel network can still be checked and corrected before a real sample is committed to it. Shortening it to save time trades a small, visible cost now against a larger, invisible one later, since a bubble left behind becomes the silent, unpredictable source of measurement error the rest of this set keeps returning to, and nothing in the eventual result will point back to priming as the cause.