I have now resolved all issues highlighted in a self-review so the only work that now needs doing here is to plug the remaining testing gaps for the new data format and chase down the pkgdown issue.

Codecov Report

Attention: Patch coverage is 98.20359% with 3 lines in your changes missing coverage. Please review.

Project coverage is 94.38%. Comparing base (d7adeac) to head (240f91f).

@@            Coverage Diff             @@
##             main     #513      +/-   ##
==========================================
+ Coverage   92.05%   94.38%   +2.32%     
==========================================
  Files          16       17       +1     
  Lines         743      837      +94     
==========================================
+ Hits          684      790     +106     
+ Misses         59       47      -12     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Unit tests are now in place covering the new functionality. Remaining small gaps are false positives from previous uncovered code.

Not evaluating the code for the failing plot section didn't help - it just shifted the error to the next plotting block. Usually this means that there is a problem upstream (which doesn't cause a error) or it could be a very oddly masked dependency issue. I don't currently see any issue locally.

Closing out all the final plotting blocks leads to the build succeeding.

The code inside the blocks that isn't plotting code works as expected on CI. This means the issue is localised within the plotting code (and repeated across both blocks).

Doing some searching and there is some suggestion this could be a cache issue (due to lazy caching being unstable with large objects). This makes some sense as the introduction of truncation here and removal of filtering the data has increased the potential size of the object quite a bit.

It looks like this was indeed a cache issue