We just got this test failure on s390x in Debian

GLib-GIO:ERROR:/«PKGBUILDDIR»/./gio/tests/socket.c:916:test_timed_wait: assertion failed (poll_duration < 112000): (167761 < 112000)
Aborted
# random seed: R02S53bb10860ffdf178b75a4913e3167afa
1..18
# Start of socket tests
ok 1 /socket/ipv4_sync
PASS: socket 1 /socket/ipv4_sync
ok 2 /socket/ipv4_async
PASS: socket 2 /socket/ipv4_async
ok 3 /socket/ipv6_sync
PASS: socket 3 /socket/ipv6_sync
ok 4 /socket/ipv6_async
PASS: socket 4 /socket/ipv6_async
ok 5 /socket/ipv6_v4mapped
PASS: socket 5 /socket/ipv6_v4mapped
ok 6 /socket/close_graceful
PASS: socket 6 /socket/close_graceful
# GLib-GIO:ERROR:/«PKGBUILDDIR»/./gio/tests/socket.c:916:test_timed_wait: assertion failed (poll_duration < 112000): (167761 < 112000)
ERROR: socket - too few tests run (expected 18, got 6)
ERROR: socket - exited with status 134 (terminated by signal 6?)

The bad line is

    start_time = g_get_monotonic_time ();
    g_socket_condition_timed_wait (client, G_IO_IN, 100000 /* 100 ms */,
                 NULL, &error);
    g_assert_error (error, G_IO_ERROR, G_IO_ERROR_TIMED_OUT);
    g_clear_error (&error);
    poll_duration = g_get_monotonic_time () - start_time;

    g_assert_cmpint (poll_duration, >=, 98000);
    g_assert_cmpint (poll_duration, <, 112000); /* here */

Does (can?) this API attempt to guarantee a maximum time or should this
assertion just be removed? The last commit (475edeb2) to this line was to allow
some leeway here - I'm afraid that we might end up simply chasing ever higher
numbers if we bump it again.