Let’s please stick to facts and keep the threads focused on what the OP asks/comments on. @free gave an honest reply of where we stand and what we’re considering for the future, I have nothing to add to this, however I want to address potentially false/imprecise accusations/observations.
The subnet in question experienced a brief spike in memory usage (barely above 450GiB) around January 24-25th (one can easily verify by the linked graphs in the linked threads). I did not follow back then and as @free explained we have limited view anyway on what each and every one canister on the subnet is doing (we mostly collect aggregates). For all we know, it was some legit spike and then it dropped (likely canisters were re-installed or so but of course that’s a guess).
The other thread you linked referred to a cycles burn spike (which was related to a large number of installs of canisters as I replied on that thread) on March 9th and there was no spike in memory usage of the subnet (again can be easily verified through the public dashboard if one expands the state chart here over the last 3 months). Therefore, there’s not even an argument that can be made about the subnet going beyond the memory reservation threshold which could affect other canisters on the subnet (by forcing cycles to be reserved for extra memory allocations). Any canisters that were “drained” during that time, were “drained” because of their own usage, not because of something else.