I did encounter issues with stability, but I believe it was a limitation of disk speed. In a monitoring system which is potentially performing multiple thousands of write and read operations every second, it will very quickly bottleneck in the disk I/O system. You need a fast disk system. We caused a hard crash that was unrecoverable because of this. In another incident, we lost a large amount of data. In this process, we decided to limit the data retention to 30 days, so the table sizes are limited, and read-write operations would suffer from less latency. The bigger your tables are, the longer it takes to seek through them to find the correct read or write position. This is an exponential factor, to my understanding.