The one library that I often keep thinking 'why doesn't this exist' about, is one which is based on different core principles than most current libraries are: a library which right form the start is designed to keep evolving for years, and which could allow users to dive really deep into one, brilliant library instead of buying new ones or switching between several libraries to get the job done.
The main libs I use all have som great qualities – and some major limitations.
Mural, which was the first lib I bought in my second round of exploring string libraries had a lot of limitations, but has improved a lot with the SSS upgrade. But it still has limitations, like eg. lack of a vibrato level between medium and molto, there's no divisi, I miss better control over the portamento – and so on. What I wrote about vibrato also applies to Berlin Strings: I often miss a level which is not as intense as their molto lib, but more intense than medium. I often like the tone and expressiveness of Afflatus, it comes with a number of section sizes, it has auto-divisi and polyphonic legato –
but I miss round robins, vibrato control (the most intense vibrato is medium-ish), the lowest and next-lowest note share the same (non-vibrato) samples. Then there's Performance Samples
Con Moto, which can sound really inspiring and expressive, but which misses most articulations, it could see some UI improvements, and doesn't have any expressed portamento at all.
And so on.
This is why many of us are investing in multiple libraries, which of course helps the manufacturers pay their bills. Creating a 'perfect library' could easily stop that – but nevertheless, I wish more library makers would try. If planned as a modular library, they could still make a lot of money on it, because it would keep expanding. Most sample library makers aren't even born yet, so it would IMO make sense to think long term, as in really long term – for a number of reasons. One main reason is that when a user has found a library s/he really likes, and then at some point discover that he's like to have a Violin 1 which is a little more expressive without being too intense, or had a rebow solution which sounded more convincing, or a different kind of portamento, s/he probably rather buy that from the same company over ending up with dealing with a lot of different UIs, sections from different libraries that don't really match each other so well - and so on.
Even when looking at attacks, rebows, legato, portamento etc limited to string libraries from only one company, they often sound quite different from each other. Sometimes they also behave quite different different between the V1 and V2, but I believe that many of us would prefer a modular monster library with instruments that were consistent, and also with some kind of minster modularity: record the instruments with, say, 8 various degrees off velocity controlled attacks, and release it with maybe three. If I could buy new and better attack options for libraries I own, I would have done that. The marcato attack is sometimes too intense, or not intense enough... and it would be great to use a system that was even more responsive to my velocities than most current libs are. There's an almost endless list of stuff that could be improved in string libraries, and as a result of that, some companies keep producing new libs. But my feeling is thatch kind of library I describe here also could generate a lot of income, especially after the initial release.