The first pattern:
re.search('(\d*)(?<=a)(\.)', '1a.')
says to find zero or more digits, followed by a dot. Right before the dot, it has a positive lookbehind, which asserts the previous character was an a. In this case, Python will match zero digits, followed by a single dot. The lookbehind fires true, because the preceding character was in fact an a.
However, the second pattern:
re.search('(\d+)(?<=a)(\.)','1a.')
matches one or more digits, followed the lookbehind and matching dot. In this case, Python is compelled to match the number 1. But then it the lookbehind must fail. Obviously, if the last character matched were a number, it cannot be the letter a. So, there is no match possible in the second case. Even if we were to remove (?<=a) from the second pattern, it would still fail because we are not accounting for the letter a.