I'm going to be honest here. Clip management is one of BI's weak points. Most of the time everything is fine, but if and when something does go wrong, BI doesn't seem to be very good at recovering on its own.
Most of the systems I work with produce such a small disk load that the disk never fills up, but I have two systems that regularly fill up the disk and I've had problems with each of them.
One system was my old BI 3.66 box I keep around for 24/7 substream recording and testing my web UI against. I logged on to that system a while ago to find two or three clumps of recordings, like the most recent 2 weeks, then a week from several months ago, and a week from even more months ago, with the clips in between all deleted correctly. Probably the cause of this system's problems is that it records to a USB external disk that sometimes (every couple of months) disappears off the system suddenly and has to be unplugged and plugged back in. It keeps the clip DB on an internal SSD that doesn't disappear randomly. My guess is that the nightly DB repair is actually just a Compact operation, and it is probably seeing the disk is gone and nuking the DB to start fresh. I don't know if this is the case, but it would explain the results I'm seeing.
The other system I had trouble on was running BI 4.4.0.0 or something, and is one I set up for someone else (using motion detection). I found it complaining about its 420 GB storage allocation being overallocated by 200 GB. When I investigated, I found that half the clips in the Stored folder were not in BI's clip list. The first thing I did was update BI. Then I manually repaired the clip DB. This corrected the issue at least temporarily. It has behaved as expected the last few weeks, deleting the old to make room for the new, but then again it behaved fine for a solid month before it screwed up the first time. It is recording to an internal drive in this case ... the OS drive actually ... so I have no clue what caused the problem. Maybe a power outage or something.