And now for something very complicated to code but very easy to use for the Anwender:
A feature to scan for candidates for non-translatables en give suggestions for regexing them.
Of course we already have this:
Defining non-translatables before you actually start translating, has the advantage that automatic placing of surrounding tags (tags surrounding these non-translatables) will be optimised. Thus saving time, increasing your productivity, in an ever changing world, where DeepL C.S. is having negative effects on the amount of offered work and prices per unit.