Re: triage: In my test environment I can demonstrate that err_network_timed_out does not draw a 'Show saved copy' button 
as far back as 50.0.2661.26
I also tested 55.0.2883.87, 57.0.2987.98 and 61.0.3163.49 (current beta)
with the same failure for page time out.

FYI: if I refresh the test page quickly (<10s) after enabling the network
I can get an err_network_changed state, which does draw the 'Show saved copy' button, but that's an edge timing use case, separate from this CR.

Here's the scenario I use to test.
-WiFi connected laptop, on a 192.168.1.* network (mask 255.255.255.0)
-network1 gateway/router 192.168.1.254, network2 gateway 192.168.1.1
-toggle network availablity with airplane mode
-while on network one visit misc router stat pages at gateway IP:
  populates cache
-close chrome session with test profile, 
-hibernate laptop, move to network2, resume windows (see Shut down below)
-open chrome test profile
-find cached URL in about:cache with network1 IP address - avoids DNS issues

Test steps (on network2)
1) airplane on, open NTP, use/refresh network1 URL, 
  get 'Show saved copy?' button
  style, adornments often missing
  using back/forward generally adds most formatting
    back goes to NTP, forward to the test URL

2) airplane off, wait 5 seconds, refresh page, page times out,
  with err_connection_timed_out, no 'Show saved copy' button
  Use wait 5 seconds to avoid err_network_changed case.

3) airplane on, refresh URL, get 'Show saved copy' button,
  verify cached still OK

Close test profile session, close chrome, 
open next test version of chrome, open test profile

I have, on occasion, demonstrated the same effects
  time-out: no show saved
  airplane: show saved
after a full laptop shutdown, restart cycle and pulled show-saved content from cache; but generally I stop/start with hibernate

The lack of full style formatting when showing the cache copy is annoying
but not essential.  I was happy just to see the data by using airplane mode.

Previews, this looks like loading page form cache button - are you the right people to triage?

FYI: Cmt#1 GW IP for network1 and network2 both as 192.168.1.254 is wrong.
The Cmt#3 config is correct. network1 GW is 192.168.1.254, 
network2 GW is 192.168.1.1 .  

When on network2, attempts to connect to the network1 GW web interface (192.168.1.254:80) for stats time out, and no Show Saved copy button is offered.

Refreshed during triage.

ryansturm: Is this in Previews code path? If not, can you change the component to UI>Browser>Offline?

I didn't see this was assigned to me. I'm marking as available on offline page node because I don't understand the bug.

ryansturm@ - I believe there are use cases where a network comm failure will result in an ERR_CONNECTION_TIMED_OUT failure, but there are cache pages available that could be displayed via 'Show Saved Copy'.  

The 192.168.1.254 vs 192.168.1.1 scenario in the description is an example.
- I was talking to the local router as:
  192.168.1.254/stats.html  - bookmark as 254-stats
- I relocated my laptop to another network where the router addr was 192.168.1.1. and 192.168.1.254 is undefined.
- When I use the bookmark 254-stats, the now active router at 1.1 does nothing and I get ERR_CONNECTION_TIMED_OUT, but the cache pages are verifiably still present.  

I was hoping to reach them through the 'Show Saved Copy' option. 

I believe the same issue applies to transient timeouts due to mobile temporary reception degradation - a variant of offline access.

This looks to be a feature request for the error page (on desktop?)

This is error page on windows. Neither Previews nor Offline can be owners, maybe NewTabPage folks know who's button that is?

This isn't part of the desktop NTP, but is displayed as part of localized_error.

Adding edwardjung@ to route to the appropriate component & owner.

+ PMs talo@ and sposner@ who may be able to help. Who's ownership would the cached copies fall under these days.

This could be distilled to wanting the cached copy button to be surfaced on other net error pages when encountering a problem. 

This issue has an owner, a component and a priority, but is still listed as untriaged or unconfirmed. By definition, this bug is triaged. Changing status to "assigned". Please reach out to me if you disagree with how I've done this.