fix(timeout): timeouts are fatal

package.json

@@ -46,7 +46,7 @@
   "dependencies": {
     "backoff-promise": "^0.0.3",
     "bluebird": "^3.5.0",
-    "quelle": "^1.0.1",
+    "quelle": "^1.0.2",
     "request": "^2.81.0",
     "sporks": "^1.0.0",
     "squadron": "^0.0.4"

scripts/enhanced-request.js

@@ -196,11 +196,19 @@ EnhancedRequest.prototype._shouldReconnect = function (err) {
       'Failed to fetch', // ECONNREFUSED/ENOTFOUND in Chrome
       'Type error', // ECONNREFUSED/ENOTFOUND in Safari
       'XHR error', // ECONNREFUSED/ENOTFOUND in Firefox
-      'EAI_AGAIN', // Transient DNS error
+      'EAI_AGAIN' // Transient DNS error
 
+      // TODO: It is not clear why CouchDB responds with these timeout errors and immediately
+      // retrying the request often leads to deadlocks when you are continuously listening. In
+      // practice, it is better that your app throws a fatal error and then is restarted, e.g. via
+      // Docker, as this seems to reliably recover from this state. At some point we should analyze
+      // this better and determine where the bug lies and fix it at the root cause:
+      // https://github.com/redgeoff/spiegel/issues/100
+      //
       // Occurs randomly even when there is a relatively small amount of data, e.g. "The request
       // could not be processed in a reasonable amount of time"
-      'timeout'
+      //
+      // 'timeout'
     ].join('|'), 'i').test(err.message);
   }
 };