Let us consider a number, e.g. 1234
Now reverse the positions of the terminal digits, so we get 4231
4231-1234=2997 which is divisible by 9
i have seen this for n-digit numbers, where n ranges from 2 to 6
Is there any number which is contradicting this behavior irrespective of its number of digits?
Now coming back to the example,
Now 2997/9=333 is a palindrome
Another example 923456781-123456789=799999992
799999992/9=88888888
Is this also independent of the number of digits of the original number?
Let us consider :
95623-35629=59994/9=6666
The largest prime factor of 6666 is 101 which is a palindrome and a prime number.
Now 6666/101=66
Again 66/11(11 being the largest prime factor of 66) equals to 6 which is a palindrome.
Another example :
923456781-123456789=799999992/9=88888888
The largest prime factor of 88888888 is 137
88888888/137=648824/101(101 being largest prime factor of 648824)=6424/73(73 being largest prime factor of 6424) this in turn equals to 88 (a palindrome)
Finally, 88/11(11 being largest prime factor of 88 and a palindrome) equals to 8 which is also a palindrome.
Now as per (http://math.stackexchange.com/questions/200835/are-there-infinitely-many-super-palindromes), will it be correct to call 88888888 a super palindrome?
Can it be possible to call a number a super palindrome if it generates non palindromes along with palindromes, provided the non-palindromes finally generate using the method of division by largest prime factor, a palindrome other than 1?
The key fact is that $9 \vert 10^a - 10^b$ for all $a$ and $b$. In particular, $9 \vert (10^k-1)$ for all $k$.
Any $n+1$ digit number is of the form $$s= a_n 10^n + a_{n-1} 10^{n-1} + \cdots + a_0$$ The new number is given by $$s_{\text{new}} = a_0 10^n + a_{n-1} 10^{n-1} + \cdots + a_{1} 10 + a_n$$ Hence, \begin{align} s - s_{\text{new}} & = a_n (10^n-1) + a_0 (1-10^n) = (10^{n}-1)(a_n-a_0) \end{align} Hence, $9 \vert (s-s_{\text{new}})$.
Note that $\vert a_n - a_0 \vert = d \in \{0,1,2,\ldots,9\}$ and $\dfrac{10^n-1}{9} = \underbrace{111\ldots 11}_{n \text{ ones}}$. Hence, if $\vert a_n - a_0 \vert = d$, then $$\vert s - s_{\text{new}} \vert = \underbrace{ddd\ldots dd}_{n \text{ d's}}$$ which is indeed more than a palindrome.
EDIT
In fact, the first half of your question is true in a more general setting. Consider a number $$s = a_n 10^n + a_{n-1} 10^{n-1} + \cdots + a_1 10 + a_0$$ Let $$s_{\text{new}} = b_n 10^n + b_{n-1} 10^{n-1} + \cdots + b_1 10 + b_0$$ where $\{b_k\}_{k=0}^n$ is a permutation of $\{a_k\}_{k=0}^n$. Then by the same argument as above we have that $9 \vert (s-s_{\text{new}})$