This is due to the fact that many processes such as write/read, replication, compaction, are organized by buckets so if you have too many buckets you may ended with too much CPU/IO consumption.
Note that this limit is usually not a problem since you can store many documents and documents types in each buckets.
Do you have a specific requirement that would need many buckets?
2- Performance & number of buckets
Based on my comment above, the short answer, is if you have less bucket it will be a little faster (since you have less CPU/IO consumption). But the reality is it depends of each use case and the type of data you store (JSON or other content, with or without views, … )