There are some issues that DuckLake may need to consider optimizing in the future

[digoal]

There are some issues that DuckLake may need to consider optimizing in the future:

• After a large number of small transactions or frequent update operations, a large number of small files may be generated, which can affect compression ratio and query performance. Does it have the capability for automatic merging? However, this isn't straightforward, as it may conflict with the time-travel feature.
• For scenarios involving heavy time-based appends, is it possible to use local buffering in the local DuckDB instance, delaying writes to the remote object storage (BLOB)?

[Mytherin]

For (1) - have a look at merging adjacent files. This is already supported.

For (2) - you can of course buffer data in a local table in a DuckDB instance manually. Using data inlining it is also possible to buffer data in the metadata catalog.