A correction often stops a behaviour immediately, which makes it feel effective. What happens afterwards is where the picture becomes more complicated.
What a punisher does and does not do
A consequence that reduces the frequency of a behaviour is a punisher by definition. Its effect is entirely subtractive.
It supplies no information about what the dog should do instead. A dog corrected for jumping knows one route to greeting is closed and has no other.
The behaviour that fills the gap is whatever the dog happens to try next, which may be barking or mouthing. Nothing in the process selects for something better.
The association is with whatever is present
Learning attaches to the whole situation, not to the behaviour the person had in mind. A dog corrected while a child is nearby may associate the child with the correction.
This is the mechanism behind corrections that make reactivity worse. A dog corrected for barking at another dog can come to predict discomfort whenever another dog appears.
The intended lesson and the learned lesson diverge easily, and there is no way to check which one occurred except by watching what the dog does later.
Why intensity tends to climb
Animals habituate to consequences delivered repeatedly at the same level, so a correction that worked initially frequently stops working.
The natural response is to increase it. That produces short-term effect and further habituation, and the cycle continues upward.
Escalation is a structural feature of the approach rather than a failure of technique, and it is a large part of why outcomes are hard to sustain.
What gets lost along with the behaviour
Warning signals are among the first things suppressed, because they are visible and easy to correct. Growling is the clearest example.
A dog that no longer growls has not become comfortable; it has learned that the warning stage carries a cost. The subsequent stage is what remains.
Reduced willingness to offer behaviour at all is another consequence. Dogs that have been corrected frequently tend to become cautious about trying anything new, which slows all later training.
Where the alternative sits
Teaching an incompatible behaviour addresses the same problem additively. A dog cannot jump while sitting, so rewarding the sit removes the jump without a correction.
Management does the rest. Changing the situation so the behaviour is difficult to perform costs nothing in trust and produces no side effects.
Behaviours involving aggression or fear are outside the scope of any self-directed approach, and corrections applied to them carry a real risk of escalation. Those cases need a qualified behaviour professional and veterinary involvement.