After a 3 months break this summer I have started working on the reproducibility project again in October.
My development work on this front is focused on exposing the columnar dependencies of operations, so that one can check the validity of a JSON history before applying it.
Given a list of operations, we should eventually be able to check which columns are referenced by each operation, and make sure that those columns are present in the project (or created by an earlier operation).
If they are not, we will give the user the opportunity to specify which columns should be used instead.
The relevant PRs on this topic are: #6740, #6929, #6974.
I have also improved the error reporting of the existing Apply dialog (#6904, #6907).
Overall, it's rather difficult to get those PRs reviewed and merged. Because @tfmorris is the only other person active in PR reviews at the moment, I am quite dependent on him to get PRs merged, which can take quite some weeks. I try to take care of reviewing all other incoming PRs myself, so as to keep the load manageable, but it's still a significant limiting factor.
We also have different expectations about the stability of various interfaces, which feels like a significant friction point.
Another activity which kept me busy in October is handling the vulnerability reports which accumulated over the summer. The series of reports we got was unprecedented in scale and it took quite some effort to mitigate them. And to fix the mitigations as the bugs in them became apparent.