Compressed buckets are currently sorted by the time field. This becomes a challenge with always compressed buckets as compression is append-only. It's common for measurements to be inserted slightly out of order. To work around this, we will have to modify the BucketCatalog to support having multiple buckets open for the same metadata.

From my initial look at this, it looks like the appropriate data structures are already in place to support this. So we'll need to add some testing around this.