Hi @Gueorguy and thank you very much for your feedback.
Regarding the open points you mentioned :
It appears that local orbital frames are always considered inertial. I have sent an email to the CCSDS office for confirmation and am still awaiting an answer. However, i am fairly confident in this.
I would say that it is simply a matter of how conservative you want to be. Probability of collision are, to begin with, based on propagated covariances which are themselves not very reliable especially if at least one object is not very well known. I would like to tackle this issue by using the work of Florian Humeau from his internship (which is doing an OrekitTalk today by the way if you are interested, link here)
I will be quoting JSpoC : “It requires that the sizes of both objects be input by the operator.” so unfortunatly we do not learn much from this. This is also because of the current CCSDS CDM blue book which do not clearly define a field for the hard body radius whereas the next version will probably have it (see most recent CCSDS CDM pink book). I would say that we will have to make do with this for now but It will be improved at some point in the future.
I agree that, in the best case scenario, we would have pre-built collision computations models following what is done by institutional screening services services models as you said.
Finally, the release of the 12.0 is getting closer and closer so i will not make major changes now. However, I would welcome feedback on the current architecture (which could be improved I think) and regarding the POC in general as you just did @Gueorguy.
Again, thank you very much for your feedback ! I will try to improve it but that will have to wait for after the 12.0.