How is the 'Relativity of simultaneity' in observing the best confirmed double-spend in a COO-less Tangle resolved?
For example, in answering this question, it was answered that the double-spend "which is directly and indirectly referenced by more transactions (i.e. has a larger cumulative weight) will most probably get confirmed."
However, the problem remains that in a COO-less Tangle, the position of the observer also plays a role in determining which spend has a larger cumulative weight.
In one node the transaction T1 might be decided to take precedence. In another node, at the same time, T2 might be decided to take precedence. On top of that, a few moments later, each node might still come to a different conclusion.
So for the understanding of participants, where a dispute could occur as to what the state of the network is, how is the problem introduced by 'Relativity of simultaneity' solved?