NSURLConnection randomly fails until device reboot - ios

Question

I have been struggling with an issue where NSURLConnection calls instantly fail. The device needs to be rebooted entirely or Flight Mode needs to be toggled on/off to resolve the problem. Restarting the app (swipe up) alone does not help.
Some facts:
-All URLs are HTTPS, TLS 1.2 compatible with Forward Secrecy. There are no issues with ATS and iOS 9. The error has been present since iOS 7 and remains on 9.2.
-No third party frameworks are used by the app. I use only native NSURLConnection calls that always work, except for when this odd situation occurs.
-No infrastructure/network issues - other devices on same networks (same WiFi for instance) work in the same app at the same time. Going to/from 3G/Wifi makes no difference.
-I always implement willCacheResponse to return nil.
-The service is hosted on AWS Elastic Beanstalk, so some suggested that it might be a DNS caching issue in case of IP address changes - this seems unlikely to me and should trigger multiple errors at once on different devices, which I have never seen.
-The method called is didFailWithError, instantaneously, as if there were no Internet connection on the device at all - all other apps work, however.
-The website that hosts the API used by the app can be browsed with no problems at all times. The website actually makes the same requests to fetch data.
The error code returned is -1003, kCFURLErrorCannotFindHost. I've been following a thread on Git dealing with the same issue to no avail. https://github.com/AFNetworking/AFNetworking/issues/967
I tried using NSURLRequestReloadIgnoringCacheData for all my requests, but that did not help.
With this information, will anyone care to venture a guess what I might be doing wrong? I added the bounty because I have no idea how to approach this problem - especially because it's so inconsistent. And it is definitely not a legitimate error (that is, that the domain could not be found), as the service is operating fine while this happens on random clients.
I create my request with a static method that looks like this. It's been stripped of some non-public info, but basically it just performs a POST request with JSON data. [Controller getSQLHost] just returns a URL - the base domain.
+(NSURLConnection*)initiatePOSTwithJSONDictionary:(NSDictionary*)dictionary toURL:(NSString*)urllocation withDelegate:delegate {
NSMutableDictionary *connectionDictionary = [[NSMutableDictionary alloc] init];
if (dictionary) {
[connectionDictionary setObject:dictionary forKey:#"input"];
}
NSData *jsonData = [NSJSONSerialization dataWithJSONObject:connectionDictionary options:kNilOptions error:nil];
NSURL *url = [NSURL URLWithString:[[Controller getSQLHost] stringByAppendingString:urllocation]];
NSString *postLength = [NSString stringWithFormat:#"%i", (int)[jsonData length]];
NSMutableURLRequest *request = [NSMutableURLRequest requestWithURL:url cachePolicy:NSURLRequestReloadIgnoringCacheData timeoutInterval:30.0];
[request setHTTPMethod:#"POST"];
[request setValue:postLength forHTTPHeaderField:#"Content-Length"];
[request setHTTPBody:jsonData];
return [[NSURLConnection alloc] initWithRequest:request delegate:delegate];
}

Answer 1

Does you delegate implement connectionShouldUseCredentialStorage ? (or respond with YES)
I think the device's keychain is used when this method returns yes which may explain the persisting failure beyond the life time of the running application and why rebooting or otherwise resetting network connectivity "fixes" it. If an authentication failure has been recognized once, it may linger in the key chain for a little while which would then respond immediately without actually going to the server.
What would cause the authentication to register as a failure in the keychain in the first place may depend on a variety of factors. It could be as simple as a typo in the saved password or more convoluted such as some certificate expiration preventing the SSL layer from establishing a secure link.

Answer 2

You're creating NSURLConnections on the current runloop. How many are active at any one time? Have you considered using an NSOperationQueue so that you don't get bitten by load bugs?
Is your delegate thread-safe? If not, that could explain the sporadic-ness of the bug.
You say you don't see the problem often, but others do. Can you borrow their devices and maybe even them and their usage patterns and thus get to see the problem more often?