that interval is too long. We need to setup script that dumps index status output with 10-30 minute interval and dumps RSS of daemon process to line up these numbers.
I also worth to counter INSERT \ REPLACE \ UPDATE \ search operations and dump these counters regular along with previous data.
With this data we might narrow down what cause memory consumption.
It case memory issue from particular search or INSERT you might create locally reproducible example.
We also might rebuild daemon with DEBUG_LEAK enabled option however daemon will be very slow this way and logs alot of data. That is why it worth to get initial investigation results then setup daemon with DEBUG_LEAK enabled option as blackhole agent at another box and check memory issue there.