I presume it is to do with the maximum blocking power of the ink. If the .quad allows you to add MORE ink than a profile for another paper then you could never get true whites (or blacks in a positive process) when using an ICC profile for a generic paper.
That's part of it, yes. Related to this is the ink configuration; e.g. for an alt. process negative you'd typically use only a few channels (or perhaps even only one), and that's what QTR allows you to do. This brings me to something else: you mention ICC and QTR profiles as if they're similar, but really, they aren't! An ICC profile is basically a color translation table or algorithm that translates between two color spaces. A QTR (.quad/.qidf) 'profile' is a translation between color data and amounts of ink to be deposited from different channels. They may seem somewhat similar from a distance, but they're entirely different animals. Note also that an ICC 'profile' for an inkjet printer is really not a deterministic profile, but rather a descriptive characterization of the printer's response to given color data. This can then be taken into account by the application software to determine an adjustment based on the printer's real-world behavior. This is fundamentally different from a QTR ink description, which really controls the actual printer behavior.
a maximum blocking ink value is a way down the scale of what could be applied by the printer.
It depends on the media and if you're talking about alt. printing processes, the process involved and even such intricacies as wavelength distribution of the light source.
Note btw that the relationship between optical blocking power/density and ink load is not as straightforward as it may seem. You might be led to believe that the blocking power of e.g. a matte black ink is sufficient to 'get the job done' and therefore there's never a reason to 'max out' the ink deposit of that channel. However, in reality, what you often do with alt. process negatives for continuous tone application is combine a couple of channels to get smooth tonality (actual resolution per channel is only a fraction of the nominal resolution of the printer). Combining several inks with varying blocking powers means that you'll easily end up deposting a lot more ink per surface area on the OHP film than the film can bear. In short, the maximum ink deposit the printer can give is in practice not the limitation you run into. It's the film's ability to take this ink without causing problems with puddling, uneven drying etc.
Or do you create corrections in Photoshop and then use those to create .quod files?
No, correction curves as used in e.g. Photoshop are distinct from QTR ink definitions. They can (and do) stack up on top of each other. Basically, the QTR ink definitions determine the printer's behavior, any additional adjustment curve can compensate for the specific response of the alt. process on the digital negative. What makes it a little confusing is that you can integrate this correction curve into the QTR ink definition.
Is it feasible to develop you own way of optimising an intermediate image and get similar results to using QTR etc?
Yes and no. In terms of linearization & adjustment curves, yes. This is what Easy Digital Negatives and similar workflows do. But what sets QTR apart is that it effectively takes the place of the printer driver and directly controls how much ink from each nozzle is being output (sort of; close enough for our understanding). So to accomplish something similar to what you can do with QTR, you'd have to do linearization in e.g. Photoshop and control the printer's behavior with the Epson printer driver. The good news is that the Epson driver offers fairly good (and in practice, sufficient) control over which channels are used (using the advanced B&W mode and its color toning function) and total ink density (through the media/paper profile, which can be manually adjusted).