Shared vs Dedicated Plans for High-Memory Bot Stability

When deploying a resource-intensive bot on Discloud, the choice between Shared and Dedicated plans involves a trade-off between cost-efficiency and resource predictability. The Shared plan operates in a multi-tenant environment where CPU and RAM are distributed, potentially introducing latency spikes due to other users on the same node.

In contrast, Dedicated plans provide isolated resources to prevent the "noisy neighbor" effect, which is critical for applications that operate close to their RAM limits. Since Discloud strictly enforces memory allocations via the OOM killer, an application that fluctuates in memory usage may face instability on a Shared plan.

Given these constraints, what are the specific performance thresholds where transitioning from a Shared to a Dedicated plan becomes necessary to avoid intermittent crashes? Does the Dedicated plan provide a guaranteed buffer against OOM kills that the Shared plan lacks?