See also PHI19. In Differential, we can predict whether "Land Revision" (or, in the general case, some similar action) will work and inform authors and reviewers that merge conflicts are present.
Some broad considerations:
- This shouldn't naively test every open revision against every change to master since that quickly becomes explosive.
- This is adjacent to other build/test concerns: e.g., "no merge conflicts" and "no merge conflicts, and tests pass on the merged change" are both reasonable things to test for.
This is basically "do a land operation, but stop before pushing and report whether it worked".