Designing the mold and drafting (drawing) the plans for it is at best one skill set; cutting the mold is another very different skill set. These are almost always done by different people or departments. Mistakes and omissions happen in both processes. Ideally, the designer/drafter should check the mold against the plans with a micrometer and calipers before it's run, but once mounted in the injection machine, it would take a major flaw to halt production and dismount, clean, and rework the mold. Adding a raised detail to the plastic requires cutting into the mold and then polishing the new recess. While expensive enough and requiring great care and skill, it's a relatively quick, simple process that might even be possible without dismounting the mold. Adding a recessed detail requires welding new material into the mold, then cutting the excess material away to leave a raised metal shape to form a recess in the plastic, possibly adding one or more new KO pins depending on the size and depth of the new shape, and polishing the shape and weld area to match the surrounding finish. This is much more time-consuming and expensive, requires dismounting the mold, and is fraught with the possibility of inadvertently damaging the mold. It can be done, but if done it will be after the initial production run is completed and the mold dismounted in due course. Someone who ran the Arii/Otaki molds damaged several of them, most of the Japanese fighters, by letting the KO pins smash into the front of the mold, and you can see the attempts to repair the molds by sloppily welding in metal and recutting it. The Monogram Bf 110G mold was even worse damaged by closing it on a tool or dropping a heavy tool in it, but it was eventually much better repaired. For such a small error as Airfix made, it's rather unlikely that they will ever correct this tab, even less likely that a smaller company like Arma Hobby will add the missing panel lines on the wing and fuselage halves of their otherwise quite good 1/48 PZL P.11c (https://modellair.blogspot.com/2021/02/48-p11c-r.html), and it's why whoever now owns the Monogram mold for the F5F-5P Panther will never return it to the fighter version.
As for the sink in your first kit that was fixed in the second, look at the gates closest to that area, It's possible Airfix made one or two ever so slightly larger so more material would get to that area. This would be the simplest fix that would give the PM more latitude during the run to speed up the cycle time - at the risk of a little bit of flash if he overdid it. The faster the mold runs, the hotter it's likely to get, which can lead to more sink there or elsewhere. Setting the machine is both a science and an art, a balancing act!
BTW, QC seldom gets involved with any of this mold work. QC will compare molded parts with the mold plans and the customer's specifications when necessary and will report any mold shifts or damage, misshapen parts, flash and short shots, gassing and raw material problems (mostly too brittle or contaminated), &c during the mold run to the Production Manager and higher management if the PM wants to continue production despite out-of-spec product. QC will sometimes, with management permission, take the problem to the customer for final approval or rejection if a large count of parts is out-of-spec - 90% of the time QC and management knows what the answer will be based on previous occurrences, so this is more a matter of ritual than business. What also happens, which is against everything QC should be responsible for, is that QC personnel will be delegated to sort through the mess and pick out defective products rather than inspect current production, leading to more cruddy products in the long run.