Welcome!
As indigo notes, cognitive assessment is not an exact science. It is, however, usually somewhat stable when obtained under reasonably standard conditions. A few notes, though: the structure of the three tests you described is not the same, so the only really comparable measure is the global measures (FSIQ), which are actually quite similar between the earlier WISC-V and the more recent WASI-2. Both of those instruments are in the Wechsler family, and thus more comparable than either is to the SB5. The SB5 has a very different design, where the VIQ and NVIQ each include measures that cut across the WISC-V VC, VSI, FR, and WM indices, and the five factor composites are composed of both verbal and nonverbal measures. This can make it quite challenging to have any kind of coherent comparison between the SB and the Wechslers.
The WASI-2 PIQ is more akin in design to the old WISC-IV PRI, and to the current WAIS-IV PRI, which both mix measures considered VSI and FRI on the WISC-V. Even without factoring in that mixture, the recent WASI-2 PIQ is pretty much identical to the old WISC-V FRI.
So the most notable difference is the drop in VCI from the WISC-V to the WASI-2. At this point, any explanation for this would be rather speculative, but FWIW, I'll offer some of that speculation:
Over the course of four years, quite a lot of influences could have lowered her normative standing, including prolonged disengagement in her own education, which may, for some learners, mean that they are immersing themselves less frequently in challenging text, which can limit growth in vocabulary and higher-level language development.
There are also, of course, the always-possible effects of regression to the mean, where a second measurement of any extreme result would be expected to become less extreme.
Another possibility is that the original postulation that her VCI was an underestimate due to the difference between her experience as a bilingual learner in another country was mistaken; it is not impossible that her education in that previous educational system gave her a higher level of verbal achievement than that obtained by an hypothetical cognitive peer operating in the US educational environment, resulting in artifactually high VCI scores. After four years of (I guess you'd have to say inferior) US education, the inflationary effects of international education have worn off, and this then might be imagined to be her "true" VCI.
Keep in mind that, in all of these scenarios, regardless of the "true" numbers, she is still a very intelligent and capable young person, with outlier instructional needs.
