Unfortunately this wouldn't actually help as much as you'd think.
It wouldn't help much with space because the bulk of data in a video stream gets produced when there is motion. Camect is already prioritizing the "interesting" video higher than uninteresting motion and thus deleting uninteresting motion more aggressively -- i.e. most of your storage space already goes to interesting video.
It also wouldn't help much with processing because of the way video is encoded. You can't start decoding a video stream at an arbitrary point in time because data coming in at a particular time is often dependent on data that came before it ... i.e. you have to be continuously processing the full resolution stream it in order to be able to start picking up data from it as soon as you determine that something is interesting. There's also the complication that by the time you process the lower-res stream and determine that something interesting happened, that event has already happened -- i.e. you need to be either holding the full-resolution video in memory or continuously writing it to disk so that it's still around when you decide you need to save it.
If you continuously recorded the full resolution video to disk without processing it, you could use the lower resolution video to know where to look in the full resolution video in the event that something happens and the low resolution video doesn't provide enough detail. Many users of 4K cameras do exactly this in order to get more cameras handled by a single Camect hub -- They hook up the cameras to both a "dumb" regular NVR recording at full resolution, and to Camect running against a substream. If the substream is high enough in resolution, (e.g. 1080p), it's good enough for what you need 95% of the time, and if you really need the full resolution video, you know from Camect exactly where to go look in the recording on the full resolution NVR.