We are considering disabling NAKs on some persistent, OTR-enabled receivers. In cases where these become slow receivers, we would like to avoid putting pressure on the source (and primary receivers) and have them switch to paced OTR recovery mode. Could you provide some feedback and guidance if we were to take this route? Are there any problems that might arise that we should keep an eye on? Any configurations you would suggest using when operating in this mode?
In some of our performance tests, we have seen some unrecoverable losses with OTR when using the default otr_request_message_timeout value of 60 seconds. We’ve extended this value out to 5 minutes and seen unrecoverable losses effectively eliminated. Are there any gotchas with extending this value? Is it liable to cause undue burden on UME stores?