We have identified a memory leak causing the issue. We are still monitoring the situation but the platform has stabilized. We will be implementing long-term fixes in the coming weeks and improving our monitoring to be able to detect this kind of issue in advance.
