Because zstd is not mature enough. Facebook claimed it was production stable, yet even with 1.5.1 it created corrupt archives at level 16 or above compression. Moreover, zstd is a single file compressor which means it needs a container for all files (like TAR) before initiating compression. I benchmark and observe zstd apart from blosc and many other projects aiming for Pareto frontier, however I find it quite concerning nobody thinks about Pareto frontier in terms of distribution of compressed contents, since compression is only performed once.

7Z is a full featured stable general purpose archival format with more than 25 years of development and track record, and has support for repairable headers, Unix/Windows precision timestamp support, far stronger compression (10-20% better ratio than zstd) and incredible ease of access (7-Zip on Windows and compress/decompress support built into Linux distributions).