Exactly, they never do. That's why calibration and validation by controlled experiments is so important.
This particular imperfection of using the same bubble model parameters for all compartments is interesting because it is the reason why VPM protects fast compartments at the expense of the slow ones, and thereby creates deep stops. Deep stops are not an inherent property of bubble models; they are caused by the particular parametrization of the bubble model.
Tissues are different regarding bubbling. Bubble growth is different in cartilage, spinal chord, or blood. The impact of bubbles on DCS symptoms is different, too: blood can endure more and larger bubbles than the spinal chord, although their halftime is not so different. All of these effects determine the allowed supersaturations of tissues, and determine which tissue should be protected to avoid DCS. A bubble model using the same surface tension and same critical radius for all compartments does not consider these differences. Whereas a Buehlmann ZHL model implicitly does consider these effects because its M-Values were determined experimentally.
This may seem paradoxical at a first glance: Although Buehlmann ZHL is mathematically much simpler than VPM and does not explicitly calculate bubble growth, it can consider the effects of bubbling on DCS more accurately because these effects were present during the experiments in which the ZHL M-Values were determined.
I think you have missed a component in the VPM there. There are tissue gas tension tracking, across an array of slowing half times - just like Buehlmann ZHL. That in itself generates the slowing of stops towards the end, and nullifies this argument that slow and fast tissue are somehow imbalanced. In the model, the Initial Critial radii you speak of above, are made into an Initial Ascent Gradients, that later becomes an Allowable Gradients (32). i.e. A slower off gas rate tissue, will retain more gas, will have a higher gradient for longer, and bubble model formula will see this and reduce the allowable gradient to keep growth within set limits. All this is part of the model implementation, and usually not discussed in the model descriptions.
What I think you should be saying, is that VPM is currently calibrated to keep the same sized micro bubble growth across the whole ascent (32 cell samples). But it also has the ability to change that to a larger Initial Critial Radii(ICR) in the early ascent (less deep stops), and smaller ICR later in the ascent (longer shallow sections). This change has not been explored yet. Maybe its time, and perhaps a VPM-D could satisfy the new trend towards the extended mid level deco.
That is a strong position to say that that bubble growth is different in different tissue types, and I suspect that it may be true also. Do you have comparative references though? I think also that most tissues are dynamic for half time, each tissue type containing a range of times. I think its very unlikely we can divided our body so neatly into 16 well timed parts. Take a look at this study about how tissue times can be influenced:
Inert gas clearance from tissue by co-currently and counter-currently arranged microvessels. - PubMed - NCBI
On the ZHL calibrations, yes of course tissue microbubble growth was present during testing, and is "baked in" to the resulting limits. But ZHL cannot create a different plan with different bubble growth conditions, by extrapolation. Bubble growth conditions and limits do not parallel the ZHL A/B limits. This is why the models diverge as the plans get bigger. ZHL is extending its lines based on its calibrated man tested data points, but like all extrapolation, that gets out of proportion after a while.
In contrast, VPM is creating new limits on each dive based on physical growth properties of micro-bubbles, all within the bounds of current tissue pressure levels.
.
