Indexer kafka crash issue fixed. Rapidjson crash was coming from the memory limit, we are now limiting the kafka load limit(ops.memory.maxrecord.count) so that it works fine with low memory pods.
Search s3 limit update issue has been fixed for now. There is a control in config (ops.index.memory.initmb) now on how-much data to cache, which can be updated based on the memory of the search pod instance. We are also working on updating our search caching scheme, to make it more elaborate.