As an example, lets look at one species considered Provisional in Hawaii, Tanimbar Corella, and one species not considered Provisional in California, Black-throated Magpie-Jay.
The corella has a few records from the Oahu, but is largely reported from the Big Island. These are all clustered in a small area around Kailua-Kona. On ebird, there are no records older than 2018, and the high count for the species in a single checklist is 7 individuals.
Now lets look at the Magpie-Jay. The population of these birds is clustered in the Tijuana River Valley south of San Diego, on the U.S.-Mexico border. On a quick perusal on ebird, the High count on any given checklist is 12, and the oldest record is 2004.
So why the difference in status? How much of this is actually down to the different attitudes towards introduced species in Hawaii vs California, the latter a former stronghold of the idea that NO introduced species should be ever be "countable".
So the Tanimbar Corella actually provides a pretty good counterexample - or if you'd rather, a reason for a "Provisional" classification is for Cornell to track populations which are identified as having potential to be Established.
While eBird records show these as being in Kona since 2018, the species is actually better found at the Lyon Arboretum and the upper Manoa Valley in Oahu, tracing to release in 1987 and a high count of at least 30 individuals. They have been reported in scattered places around Oahu, as well as Maui and Hawaii, as you've noted - so there is at least the possibility that unless these are repeated releases, they may be spreading. Taninbar Corella is a relatively recently described species and among many is better known by its trade name - Goffin's Cockatoo - so that complicates records a bit.
When I asked the California BRC about Black-throated Magpie-Jay, I was told that while this was a borderline case, the CBRC has been tracking this species for decades, it is currently "hanging by a thread" with recent CBC counts of less than 5 in the core of its range, which has not expanded beyond the Tijuana River Valley. Personally I wonder if it has contracted due to the same winter events which got the waxbills.
Taking both of these into account, I think some lessons may be distilled:
1) The value of local knowledge in setting these categories. In both of these examples, the true sense of these populations is not available from a simple scan of eBird. We in the peanut gallery may be drawing our own initial assessments from afar, but we are not dealing with a complete picture until we get information from the people actually studying these locations.
2) The value of incentive for eBirders. Maybe the corella reports in eBird are buried under "Exotic" status... or maybe they are just not being reported by birders who don't consider them a "tick." It may not be explicitly stated, but if Provisional species are to "count" in eBird's system, then some observers will pay more attention to them. The birder/researcher dynamic of eBird is an interesting one of different incentives. I'm sure the research-oriented folks at Cornell would care very little about providing venue for people's personal checklists or features such as "Top 100" except that it incentivizes people to submit vast amounts of data. I don't think this incentive is to be underestimated - there is a great deal of talk regarding Clements/eBird as a competing (even displacing) taxonomy and listing system as compared to AOS and ABA especially. Such talk only occurs if people are taking it seriously. If the corella is given "Exotic" status and thus is not "countable" (as, in fact, it has been for decades in this and other systems), then the important trends get lost due to lack of reporting, as I mentioned in point 1. This does beg the question, then why not track any breeding species, except:
3) The focus seems to be on expanding or stable populations rather than contracting ones, and if someone wants to draw a line between Exotic and Provisional, this might be where it is. Its an arguable point and this is where some inconsistency can certainly creep in - the magpie-jays were certainly multiplying if not geographically expanding for a number of years. Are they Provisional in certain years and not in others? What of Crested Mynas in Vancouver, then? There aren't easy answers. I think Cornell is relying on that local knowledge to course correct when needed. But I would argue that this imperfect situation reflects the imperfect knowledge we have of avian population dynamics - sometimes the correct answer is that we don't know if something is "Established" or not. A system with no middle ground leaves no room for this. I'll add that I think a middle category complicates the matter no more than already exists. As we've demonstrated, people already argue over whether species are established or not, and which committees have it right and wrong and so on. At least with this system, there is something between ABA-level "countability" and a run of the mill cagebird/waterfowl escape - a situation which I think reflects reality.
