salted_sha_pw_enc() provides this much space for the hash value:
    char hash[ MAX_SHA_HASH_SIZE + SHA_SALT_LENGTH ];
that is, it wants hash to contain the hashed value + the salt.  But when it actually generates the hash
    /* hash the user's key */
    if ( sha_salted_hash( hash, pwd, &saltval, secOID ) != SECSuccess ) {
        return( NULL );
    }
the length of hash is always the length the hash algorithm produces - SHA1_LENGTH, SHA256_LENGTH, etc. - even though salt is passed in, and added to the hash value, it is not appended to the hash.  salted_sha_pw_enc() allocates space for the password hash plus the salt
    if (( enc = slapi_ch_malloc( 3 + schemeNameLen +
        LDIF_BASE64_LEN(shaLen + SHA_SALT_LENGTH))) == NULL ) {
        return( NULL );
    }
however, it never actually writes the salt value into enc - it only writes the first shaLen bytes from hash, plus SHA_SALT_LENGTH bytes of garbage.  I think the assumption was that calling sha_salted_hash() would append the salt value to the end of hash, but this is not done.  I think the fix is to have sha_salted_hash() do something like this, after generating the hash:
    memcpy(hash_out + outLen, salt.bv_val, salt.bv_len);