If you're suspecting resource leakage, you could check is processes memory usage with "smem" tool (PSS) column (in smem package).
If that doesn't find out the problem, start taking regular snapshots with this tool: https://maemo.gitorious.org/maemo-tools/sp-endurance
It has separate script for taking snapshots of all resource usage information in the device, and couple of scripts for doing from those snapshots HTML reports & drawing graphs of all possible resources that could be leaked: process & kernel memory, shared & locked memory, file descriptors, CPU, network and other IO, errors in syslog (mostly relevant for phones) etc.
The snapshots can be very large, depending on your syslog size, how many processes you're running etc, so I would recommend using large enough snapshot internal that when the problem starts to show up badly, you have something like ~20 snapshots. That should be enough for showing whether there're any kind of resource leakage trends.
(If you want to package the tool, RPM & Debian packaging files are in separate branch in git, but at least Debian packaging is a bit behind master. If you're interested, I can provide a patch to apply to packaging after merging master to packaging branch.)