re 1. That the preview in the review-system does not work is annoying but known, see MathJax in suggested edits A workaround is to select 'improve' where it does work, and then to 'cancel' the improvement to then 'approve' (assuming of course one just wants to approve).
re 2. The question got edited/improved by another reviewer in parallel. The version you got for improving/editing, is the version that already contains their changes.