in SQL Server 2005+, they no longer drop all temp tables from the tempDB. A small cache of 16 temp tables are saved in a special area of cache. This was done to prevent the overhead of constantly allocating new tables for systems that create a lot of temp tables. If an unused table exists in cache, it will use one of them, if not, it creates a new one.

This is why you will see temp tables hang around. You can read more about that on Paul Randal's blog: Misconceptions around TF-1118

If you are experiencing tempDB contention, then you likely do not have your tempDB configured per best practices. Your data files should all be the same size, and you should have multiple data file, somewhere from 1/4 to 1 data file per logical processor. I would suggest starting with 1/4 to 1/2 data file per CPU and increase it if contention persists. You should also separate your tempDB files to a dedicated drive separate from the other databases' files.

When you are experiencing heavy blocking in the tempDB, you can use the script on my blog for determining if the contention is on allocation pages (GAM, SGAM, PFS) or data pages: Breaking Down TempDB Contention