But I do not like algorithmes who count on the small (~0) chance to overrun the lower or the upper limits.
That's a very wise option.
the whole task seems to be ill posed and in need for nobody. You may speculate of or produce a ton of error and useless solutions. I prefer... to do neither