Issue #263, opened quite some time ago by @coatless, demonstrated that we had a bug stacked away in dealing with a field exporter which dropped a dimension. I worked a bit on that in November and committed a fix albeit behind a #define. I had meant to come back to this and had time last week to run two sets of 'before' and 'after' reverse-depends checks to assess how costly it would be to enable this unconditionally.

It turns out that four (out of 950 total (!!)) packages then fail tests:

• AlphaSimR by @gaynorr et al -- already addressed in version 1.1.2 on CRAN
• carat by @yexiaoqingruc et al
• odpc by @esmucler -- will revert to old behavior, see #3, #4, addressed in version 2.0.5 on CRAN
• perccal by Dan McCarthy (not on GH, emailed 2022-02-13)

I have not yet had time to look into patches or changes. We probably want to fix this, and worst case can offer a #define to allow these four packages (and other uses like them) to get the old behavior back (in the sense of 'we published an API so we need to honor that contract') but otherwise offer 'better' code.

So "just" opening this issue to see if we can start a dialogue about what would be a suitable timeline to make the change. Maybe in three months? Too soon?