Fixes in 
https://github.com/openshift/ose/pull/215
https://github.com/openshift/origin/pull/8810

https://github.com/kubernetes/kubernetes/pull/25369

Could you give me some suggestions on how to verify this bug? Thanks.

The steps in https://github.com/kubernetes/kubernetes/issues/25151#issue-153080519 should be good... request a watch on a version that is too old, and ensure the 410 Gone error is returned as a watch error event, rather than as a HTTP 410 response.

To make testing easier, set the watch cache size to something low for a particular resource type, and create enough of that resource type to exceed the cache size, then request resourceVersion=1

Test on openshift v3.2.1.1-1-g33fa4ea

Steps to verfify:
1. enable watch-cache
kubernetesMasterConfig:
  apiServerArguments:
    watch-cache: ["true"]
    watch-cache-sizes: ["builds#50","deploymentconfigs#50"]

2. watch a old resource by curl, should return a Gone 410 event.
    [root@dhcp-128-7 Desktop]# curl -k -vvv -H "Authorization: Bearer 82z8aFmWWHrzBH8-nwPPgRxs2sLepbw0re75hqaJgTs" "https://104.197.173.141:8443/oapi/v1/namespaces/dma/builds?watch=1&resourceVersion=1"
    * About to connect() to 104.197.173.141 port 8443 (#0)
    *   Trying 104.197.173.141... connected
    * Connected to 104.197.173.141 (104.197.173.141) port 8443 (#0)
    * Initializing NSS with certpath: sql:/etc/pki/nssdb
    * warning: ignoring value of ssl.verifyhost
    * skipping SSL peer certificate verification
    * NSS: client certificate not found (nickname not specified)
    * SSL connection using TLS_RSA_WITH_AES_128_CBC_SHA
    * Server certificate:
    *       subject: CN=10.240.0.29
    *       start date: Jun 12 03:16:37 2016 GMT
    *       expire date: Jun 12 03:16:38 2018 GMT
    *       common name: 10.240.0.29
    *       issuer: CN=openshift-signer@1465701391
    > GET /oapi/v1/namespaces/dma/builds?watch=1&resourceVersion=1 HTTP/1.1
    > User-Agent: curl/7.19.7 (x86_64-redhat-linux-gnu) libcurl/7.19.7 NSS/3.14.3.0 zlib/1.2.3 libidn/1.18 libssh2/1.4.2
    > Host: 104.197.173.141:8443
    > Accept: */*
    > Authorization: Bearer 82z8aFmWWHrzBH8-nwPPgRxs2sLepbw0re75hqaJgTs
    >
    < HTTP/1.1 200 OK
    < Cache-Control: no-store
    < Transfer-Encoding: chunked
    < Date: Sun, 12 Jun 2016 05:40:07 GMT
    < Content-Type: text/plain; charset=utf-8
    < Transfer-Encoding: chunked
    <
    {"type":"ERROR","object":{"kind":"Status","apiVersion":"v1","metadata":{},"status":"Failure","message":"too old resource version: 1 (1046)","reason":"Gone","code":410}}
    * Connection #0 to host 104.197.173.141 left intact
    * Closing connection #0

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2016:1343