It depends on the age at which the person takes the test. People's personality in the general sense of the word, certainly does evolve through different stages of development. I think in early adolescence e.g. like 12-16, the hormones along with just been a teen means there can be a change in personality. For instance, the first MBTI test I took when I was 16 typed me as INTJ, it was just something I had seen on a forum unrelated to typology, took it, identified with the result and forgot about it. Three years later, my friend introduces me to MBTI, and I had completely forgotten I was 'once an INTJ', until I was randomly going through my post history on the forum one night.
While Ti has always played a major role in my consciousness, I think dimensionally (Keirsey style), I was far more J than I am now, and my behavior was more in line with INTJ than TP. As I was finding myself, I think Ni temporarily took the front seat of my consciousness, and the role of Ti was more in line with a model 8 function. My areas of interest at the time were also more Te focused than Ti, e.g. rather than reading books, I was doing experiments with chemistry.
When I was a child, I was an extremely strongly typed INTP, but from like 13-16, I was tending towards being more J, and then as I was 17-18, there was a much more rapid shift towards INTP. I think for the most part, we have one, and only one true type, but it's by no means unusual for people to temporarily change into 'someone else' during the teenage years, and such a change could be detected psychometrically. Tests also can be very unreliable because they are limited by how you interpret the questions and one's level of self awareness, so the best way, is to build your own archetype of each of the 16 types and find on a personal level what's the best fit. The most egregious example I can think of an MBTI questionnaire producing a mistype was my dad whose a very strongly typed and avoidant INFP, but according to 16personalities.com, he's an ENFJ.
Probably the most fragile of the 4 dimensions is the J-P dimension because it's usually inferred through indirect questions that people often are unreliable in answering, e.g. "work then play, or play then work". By far there is the greatest error rate (in my opinion and experience) with this functionally critical dimension; I would be curious if there are any online tests that determine type by identifying primary functions. Less frequent but still very common is mistyping in the N-S dimension e.g. a sensor may be unaware of their 'lack of' intuition.
So a big factor for people's types changing on tests is their cognitive adaption towards answering questions, the more psychometric questionnaires you do, the more tricky questions you come across and thus gain greater self awareness. If your ability to accurately answer questions and your ability to identify the correctness of the result changes, then you may consistently get a new type despite previously identifying with a different type.
I've had a lot of coffee so I'm going to stop now, just in case my reply is like hundreds of words long haha