Talked to Eugene and verified that this test is already broken on 10.3, so not specifically related to iOS 11 change, and it's a glue issue. I'm disabling this test on iOS 11 and assigning it to Eugene for further investigations.

"cacheControl: max-age=0" http request header is not sent by WKWebView, which is a web compat problem.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/00fa58da78a66cbdf36820f415b437be94d6184c

commit 00fa58da78a66cbdf36820f415b437be94d6184c
Author: Yuke Liao <liaoyuke@chromium.org>
Date: Wed Jul 26 23:57:31 2017

Disable testCachingBehaviorOnNavigateBackAndPageReload on iOS 10.3.1

This test fails consistently on iOS 10.3.1 and afterwards.

Bug: 747436
Change-Id: I9c59a16cb359fb42edd47c1b1bebdfed8200dbb4
Reviewed-on: https://chromium-review.googlesource.com/588050
Reviewed-by: Eugene But <eugenebut@chromium.org>
Commit-Queue: Yuke Liao <liaoyuke@chromium.org>
Cr-Commit-Position: refs/heads/master@{#489802}
[modify] https://crrev.com/00fa58da78a66cbdf36820f415b437be94d6184c/ios/chrome/browser/web/cache_egtest.mm

Still failing on iOS 11. Danyao, do you think this is a bug or WAI?

Not sure. Safari sends "Cache-Control: no-cache" header instead of the "Cache-Control: max-age=0" header that Blink and Firefox use. I'm not sure what is the right behavior. 

Why did Bling decide to check the header in this test? The request counter already tests that the server is hit when not using cache (which I assume is browser cache). Are we perhaps over-constraining the behavior here?

+eugenebut@ for the question in Comment #6

WKWebView used to send "Cache-Control: max-age=0" prior to iOS 10.3. This test was written long time ago when Chrome for iOS used own network stack. Back then I guess it was important to test that correct headers are sent.