What problem does this feature solve?

复现demo：https://codesandbox.io/p/sandbox/vtable-performance-forked-kckdzw

普通单元格更新有相应的优化，但是合并单元格是硬更新（生成新节点）所以造成的性能消耗要大很多。希望优化

What does the proposed API look like?

与api无关