So, my first post here! I think we are far from a plateau from a technical perspective. But we might be reaching a plateau from a business perspective. Pure sample-based VST are hitting a technical limit. These VSTs resort to increase elements such as number of layers, type of samples, and sample quality. But the quality gains are becoming marginal. This is the same as discussing the actual differences in sound quality between FLAC and 320 kbps MP3.
Model-based VSTs such as Pianoteq continue to improve and, despite its current tonal limitations, they already offer better playability and flexibility than sample-based VSTs. Some sample-based VSTs also use modelling, but the results are still very basic. One of the reasons is that sample-based VSTs use quite similar engines that were not designed to combine samples with modelling. So, new sample-based VSTs are nothing but newly recorded samples that rely on the same old sample-engine technology.
Combining both technologies could be a way forward, but that would require a major change in business strategy of the VST developers. Moreover, the piano VST world is a very small niche, and what we have today is sufficient for most music production and studios. And what most of us are looking for is a even a smaller niche inside that niche, which is a VST to replace the sound engine of the DP, not to be used in music production, where sound quality is critical but playability is not. My bet is that sample-based VSTs will only react when new versions of Pianoteq start taking their customers… but this will take time.