Ok, but we also need to ensure that checksums are output in the same way as in other places in code.
For example, look at CompressedReadBufferBase.cpp:

getHexUIntLowercase(expected_checksum.first) + getHexUIntLowercase(expected_checksum.second)

But if we look deeper in code, we see two inline functions

inline uint64 Uint128Low64(const uint128& x) { return x.first; }
inline uint64 Uint128High64(const uint128& x) { return x.second; }

I used them, but missed the order. I'll rewrite my code without them. Also I used LIMIT in test script, that's why random rows were chosen and some tests failed.

But there is also a difference in lower/upper case.