The use of delete markers, also known as tombstones, in storage systems based on LSM1/MVCC can lead to scan performance degradation when the number of tombstones increases. This is because the query executor has to seek through the tombstones to find the earliest item, which can be time-consuming. To overcome this issue, a simple trick is to help the executor figure out an approximate point to start its search by using the last seen timestamp (`<last_ts>`) and modifying the query slightly. This approach reduces the number of seeks and significantly improves scan performance. Additionally, regular compaction can remove tombstones and prevent issues. However, in some cases, it may not be feasible to modify the application or handle scattered delete tombstones.