It also notes that "On occasion, timeouts will occur that are outside of the control of your script or our servers. Typical reasons for these timeouts are Internet traffic, merchant server overload or malfunctions, or Internet routing issues. Depending upon your server location and what route is used to send data, it is possible that you may occasionally receive a time out message."
It appears that Authorize.net does not retry a failed POST, even if the 10 second timeout has not been reached. This was confirmed by an admin in the forums ("We currently do not retry failed posts").
I propose that this behavior be changed. If an Authorize.net POST request fails, prior to the 10 second cut-off, the POST should be retried, possibly with a short backoff (e.g., wait a second or two to reinitiate, to prevent a flood of requests).
As background, we have been using DPM successfully for a couple of years now, but we do occassionally see "timeout" errors. Crucially, it does not appear that these are actually caused by timeouts. The first thing we do in handling the response is log receipt of the request. But we see no evidence of having received the requests in our logs. Which suggests that the problem is happening outside of our network.
As it currently stands, Authorize.net's POST request could fail immediately due to some extremely transitory issue (perhaps even within their network). They would immediately receive a "connection reset by peer" error or whatever. And even though virtually none of the 10 second timeout period has been consumed, the customer receives a timeout error.
The DPM process should make more of an effort to communicate the transaction status and prevent this failure scenario.
Possibly related to this request would be additional logging facilities, so that both Authorize.net and its customers could have more insight into what exactly is occuring. IOW, it would be very helpful to have some visibility into *why* Authorize.net's POST request failed, and how long it took. It could provide much needed stats to discover how often the "timeout" problem is happening and whether these suggested changes are actually making a difference.