Realize also that a normal population of millions of HIV virions have a huge percentage of those virions inactive. This is caused by the crash error effect: too many bad copies. With this strategy, the virus accomplishes two things: it explores the maximum possible range of mutations, so that it looks forever different to the immune system, and it saturates the immune system with defective virions and viral antigens, diverging resources and allowing for a very tiny subset (don't recall numbers, but was well under 0.01%) to escape detection.
Another key concept is the synergy of viral particles. Many of those non-crashed virions are actually only partially crashed. When they coinfect a cell, betwen both virions they have enough good proteins that they can survive; i.e. reproduce and spread. Considering the cell infection rate is extremely low compared to the viral load in blood, and the even more extremely low probability of coinfection of two complementary virions, realize that was the number of ineffective virions be increased by a tiny amount, the virus would not be able to spread any further. By all means these drugs that push RNA viruses to the crash-error zone are a very good idea.
PS: There are 10's of million's of people with HIV, so randomly mutating trillions strands of viral HIV inside people as they are reproducing and not creating a worse strain seems unlikely.