The biggest problem with RA 2.0 now is that allows infinite number of tries to enter the password .
In last few months I have noticed a lot of 4899 port scans trying to find RA and break in using simple brute force (on good network connection and with reasonable powerful computer it takes 2-3h to gues RA 2.0 password).
Same problem exists when using NT security - all that you need is user name (for example from e-mail address) and this way you can not only access computer by remote but have user password too.
So I believe 3.0 absolutely need ban timeout , something like if user entered 3 times in the row wrong password then ban his IP for let's say 15 minutes.
Also add let's say 10sec delay before passwords asking retry, this will be hardly noticable by real user that made a mistake but will slow down brute force cracking to become completely useless.
Until this implemented RA is a high level security threat.