It is like memory, which is set or reset.
When brass yields, as in up-sized during firing, it springs back to a new dimension. Same with down-size yielding. Once cases reach the interference fit where shoulder bumping is needed, they will usually have to be bumped every cycle from that point onward. This, because the brass is then more inclined to go back to where it's been (it's memory) on every successive firing. Same with brass everywhere else.
It makes sense to try & manage the bump to desired head spacing -before setting a memory from interference fit. Then it also makes sense to do this with the case bodies and necks, and custom FL dies that are local chamber associated work to allow consistent extraction.
There is another way to do the same, while BR shooters resist the notion of it.
You can go 'fitted' chamber to new brass dimensions plus no more than 1thou. With this, by never allowing new brass yielding to begin with, larger dimension memory is mitigated.
For case bodies and necks this is easy, especially with low taper case designs.
For shoulders it's a lot more difficult, because of their low and less than fully formed angles (with new brass).
Perhaps you could sneak into perfect memory while forming greatly improved shoulders, 40deg+, and then remove all sizing, all together, from that point onward.
I've already tested this with bodies & necks using a BAT RS action/26wssm Imp. It works great. But 35deg shoulders still require bumping -eventually & onward. Currently using a custom bump die for this.
I do not yet know what shoulder angle could mitigate a need for bumping. Will try either 45 or 60deg with my next chamber(haven't decided), in hopes of removing any need for resizing dies.
I figure minimum sizing is better than maximum sizing, and that zero sizing as load developed, will be better still. True or not, at least I'll learn about this much.