fukang wang

@mdellanoce I considered that this PR was not the optimal solution, and the implementation was not very elegant, so I changed it to a draft. When I am not busy...

@mdellanoce I know why "without shuffle fix" is faster than "with". The above Dom structure can be created in many ways. 1. Add `div1` as a whole to body ```js...

@wensiyuanseven 能提供个可以复现问题的仓库不，我看一下

> Me too > > Version v2.0.0-alpha.10 @chowkings can you provide a minimum git repository?

@mdellanoce Thanks for your excellent work and it worked for me to reproduce the issue. I am wondering if there is a way to optimize without affecting the execution performance...

@mdellanoce Hi Michael, I opened a new PR(https://github.com/rrweb-io/rrweb/pull/1302) without affecting the execution performance of the average case. Is it working for you?

The root cause is that if you add a node to a DoubleLinkedList, due to random sort and shuffle, the previous sibling and next sibling of this node have not...

Another interesting is that if you update the benchmark test to: ```javascript setTimeout(() => { // re-shuffle the options options.sort(() => Math.random() - 0.5); for (var o of options) {...

@mdellanoce I got 800+ms with `defragment` in this benchmark test, but why are 173.4ms in yours ```diff +public defragment() { + let current = this.head; + while (current) { +...

@mdellanoce I considered this solution when optimizing recording performance in May this year: https://github.com/wfk007/rrweb/commit/05c77713324b37781bccf1d17be7faa7f04b653e, and communicated with @YunFeng0817, but gave up **because the impact was too huge**, so I finally...