Us assigning letters to the triangles is arbitrary, as is the ordering of the triangles (i.e. we can assign 1,2,3 to a,b,d in 6 different ways). Assuming a, b, and d are interchangeable, that leaves 6 ways to assign A-Z to the pairs like:


For frequency analysis, none of this matters though, since each of the pairs is replaced by a single letter. In that case, only the mapping between pairs and letters matters, as well as the order in which the result is read from the grid (left-right, up-down, spiral, etc). So, if each pair stands for a single unique letter, we should be able to break it by reading out the pairs from the grid in some logical way, and then attack the result as a substitution cipher with frequency analysis.  Using the above substitution:


Haven't had any luck attacking this as a substitution cipher. Frequency analysis will fail though if we are using the wrong way to read out the pairs from the grid, or if we only need a subset of them (e.g. only the pairs that are attached to the dark green pipes).

Another direction would be that the aztec code is the result of a superposition of multiple aztec codes, and it can be demultiplexed by the markings. However, then the whole core of the aztec code should have the same marking.

A final thought on "the journey inward": the map seems to consist of a number of layers on top of each other: QR on top of AZTEC on top of DARK GREEN PIPES on top of LIGHT GREEN PIPES on top of a RED GRID. So maybe Level 1 of Days Destroyed would only need the QR, level 2 only the AZTEC, and so on.