Use larger interval for pollfd test

[nalimilan]

This prevents frequent failures from happening on build VMs.
It is not clear whether the failure reflects an underlying
bug or not, but having the CI fail randomly doesn't help fixing it.

Fixes or works around #12824

[ViralBShah]

cc @amitmurthy

[vtjnash]

this change isn't related to the cause of the bug (which is a discrepancy in the elapsed-time estimation), so it likely won't affect the failure rate

[amitmurthy]

Agree with Jameson. Comment out the test since we anyway have #12824 open and tracking the bug ?

[nalimilan]

I'm not sure, but 1) the failure only happens with the 0.002 interval, not the larger ones, and 2) I haven't seen the failure after applying the change. Isn't it likely that the discrepancy only happens for very short intervals?

I'm OK with disabling it too, though.

[amitmurthy]

OK. Can try it out and change the title of #12824 to something like "pollfd interval test fails for small intervals"

[nalimilan]

I've adjusted the title. Let's see whether we see this failure again or not in the next few weeks.

[nalimilan]

You were right, increasing the interval only made the failures less frequent, but they haven't disappeared. I've just seen one with 0.005 <= 0.004910521.

In PR #14074, I comment out the test as you suggested.

[nalimilan]

(FWIW, though it hasn't fixed them, it seems this already reduced the failure rate quite noticeably.)