Almost all testbenches should be self-checking and have the following minimum functionality:

• Logging and verbosity control – to allow good progress reports and simpler debugging
• Alert handling – to clearly show exceptions, count them and allow various actions
• Extended string handling – to make it easier to make good log and alert messages
• Simple randomisation – because this often detects problems you didn’t think of
• Value Checkers – as checking signal outputs is a major activity in any testbench
• Signal Expectors – as you are very often waiting for a signal (e.g. an interrupt)
• Stability checkers – as you must make sure there has been no activity or spike on a given signal 

In addition to various support to make it easier to make good BFMs (Bus Functional Models) 
All of this helps you make a better testbench faster, and improves overview, readability, maintainability, and reuse.