As far as I know the only thing LFS does on the HD during a race is periodically (every time 4096 bytes of replay data have accumulated) write the temp_spr.spr, respectively its mpr counterpart.
Now, what would happen if somehow those 4KB are piling up quicker than the HD can write? My guess is that LFS would grind to a halt until the writing can catch up again, since I doubt LFS has any checks in regard to how
long writing this data takes. If the file is somehow blocked or LFS' file handle forcefully removed, then as I've seen LFS just stops writing the file and later claims that "Replay was not being recorded," which is fine behaviour in that regard, so an outright file block doesn't seem to be the problem.
The question now is, what causes the HD to be slower than LFS requires? I can see two possibilities:
- External influence. The HD is fragmented / broken / used by another program / whatever and needs to do other stuff to fix the situation and become fast enough again.
.
- Something in LFS causes massive amounts of replay data to be generated. Maybe the steer jittering resulting from an engaged ABS somehow plays into this? It happening as mentioned for example in Blackwood T1 could be caused by lots of people braking at that point. Or maybe it's simply a bug (in the ABS code?) that causes the data flood.
- A combination of the above.
This would explain why watching the replay doesn't reproduce the behaviour, since obviously there's no replay being written while watching one. Maybe the people who are affected by this could turn off the automatic replay storing (Options > Game > ...replay save) to see if it has any effect - it should at least help narrowing down the source of the problem.