Understood. There's obviously a trade-off which needs to be considered. I believe there are strong arguments in favor of zstd, mainly speed. But if it's prohibitive in other areas, we must reconsider the decision. One obvious possiblility would be to split off kernel-default-extra like we do in Leap, reducing the install size of the kernel base package. But that's yet another trade-off, and doing it would cost substantial effort.