Looking at versions.txt 1.88 mentions "Tightened up collision detection as per bug report postings." so 1.87 may be a good version to test against.
I remember a lot of discussion about "grazing" issues at the time.
Another new behavior that is possibly causing you probs is multiple collisions being reported for a single crash. You may want to have a look at your collision handling, there is a possibility that small tweak collisions get reported as well as the main incident now and your code may be interpreting them as massive car stopping events.
I'll dig up some of the old tests if you're completely stuck (excuse the pun) to try and narrow it down if it looks like a definite fault with blitz3d .
I remember a lot of discussion about "grazing" issues at the time.
Another new behavior that is possibly causing you probs is multiple collisions being reported for a single crash. You may want to have a look at your collision handling, there is a possibility that small tweak collisions get reported as well as the main incident now and your code may be interpreting them as massive car stopping events.
I'll dig up some of the old tests if you're completely stuck (excuse the pun) to try and narrow it down if it looks like a definite fault with blitz3d .