It seems, my half-baked solution slipped through tests due to the fact that at that moment random tests was formed without any upper palindromic boundary values.
You can provide tests in which at least one of the boundariesis guaranteed to be a palindrome itself: (111, 2222), (123, 3333), (111, 2345) and nothing will accidentally slip through. )

Haha, it's OK! It finally dawned on me what it was all about and I’ve already optimized even my final solution, but I won’t republish it - I’d rather celebrate it with a couple of glasses of cognac - no one can drink away the talent for being a little stupid! Cheers!

Not easy but incredible fun - I couldn't asleep a couple of nights :D
Thanks for the great kata!
PS: BTW, is it possible to remove the fisrt debug version of my solution, that didn't pass all the tests and has been unecspectadly published by clicking on the button, that was changed from 'attempt' to 'submit' at that moment?

Thank you friend! You rlly saved me from complete ASS madness :D

Vote for 5

The initial Kata input constraints do not require this, however, here is a solution in a more general form, which covers (it seems) all the expanded cases with the long digits and unnested parenthesis:

solve("121(a2(b)a)")
solve("33(A2(B)A)")
solve("k(a3(b(a202(c))))")
solve("a2(b)c2(d)")
solve("a2(b)c14(d)")
and even with zero:
solve("a2(b)3(c0(demonicstring))"

Cool Kata, thanks for the fun!