You need to add an entry for the ACS server into the LMS server's local hosts file. that will correct the performance problem. If pieces of HUM were not working with ACS integration, then I imagine you did not integrate correctly. Make sure that your System Identity User and your login user both have the necessary tasks rights and device access in ACS. For the System Identity User, you must grant all access for HUM for all devices.
Note: there is also a bug in HUM 1.0 if the System Identity User is not admin where jobs can fail to work. The bug is CSCsr93292, and is fixed in 1.0.2 (available from http://www.cisco.com/cgi-bin/tablebuild.pl/cw2000-hum ).
Local authentication is not meant to work when ACS is unreachable. In that case, you have maybe one fallback user (configured in the TACACS+ login module) who can login, and make emergency changes. Typically, though, the solution is to [temporarily] set the login module back to local using the NMSROOT/bin/ResetLoginModule.pl script. A better alternative is to setup multiple ACS servers for redundancy.