Scanf not reading piped input's first value only as expected (reads as string, expected integer)

There is no such thing as relevant code... the problem might lie in lines that you did not post. Yet there are multiple issues with the posted fragments:

• the return value of scanf() is not tested so one cannot detect conversion errors (actually you document the return values, but I suspect you get them from the debugger).

• it is very difficult to recover from conversion errors when parsing with scanf directly. One should read the input one line at a time with fgets(), parse it with sscanf() and test the return value to detect errors and report the errors with meaningful messages.

• prog1 use %zu.%-6zu to output event1.time.tv_sec and event1.time.tv_usec. This poses 2 different problems:

  • the types of tv_sec and tv_usec are not size_t, they are system dependent types time_t and suseconds_t. You should use %lu or %llu and convert the fields appropriately. As coded, the program may have undefined behavior.
  • using %zu.%-6zu to output a decimal value to be parsed with %lf is incorrect because the microseconds are not always output with 6 digits and smaller microsecond numbers will produce fewer decimals that will be parsed with an incorrect value by %lf. You should use %lu.%.6lu or %lu.%06lu instead. As coded, 1 second and 5 microseconds is printed as 1.5 and read back as 1.5 instead of 1.000005.

Here is a conversion loop that should make the parsing errors easier to analyze:

#include <stdio.h>

int main() {
    int event_num, type, code, value;
    double time;
    char buf[100];
    while (fgets(buf, sizeof buf, stdin)) {
        int res = sscanf(buf, "%d %lf %d %d %d",
                         &event_num, &time, &type, &code, &value);
        if (res != 5) {
            fprintf(stderr, "parsing error, returns %d: %s\n", res, buf);
            fprintf(stderr, "  initial bytes: %02hhx %02hhx %02hhx\n",
                    buf[0], buf[1], buf[2]);
        } else {
            /* handle the parsed values */
        }
    }
    return 0;
}

You could also check for potentially invalid trailing data at the end of the line with an extra %c or %n conversion.

Here is a wild guess: you might want to check the output stream for an initial byte sequence EF BB BF using a hexdump utility. This sequence is a UTF-8 encoded Byte Order Mark that may have been inserted at the beginning of the output by prog1 depending on the locale settings.