The rule of thumb for batch.opensrs.net is as follows:
There are no numerical rate limits as to how often or how many connections a reseller can make to the batch pool. However this system is monitored for performance as well all API rate limits are enforced in the same manner as the LIVE system at the application level. If an IP and its related process requests, causes the system to suffer performance issues that negatively affect all batch users, then the offending IP will be blocked for a period of 24 hours and automatically reinstated the next business day.
That sounds more like a rule of thumb-up-your-collective-butts to me.
You really need to set a standard of acceptable use, then configure sufficient hardware to accomodate it.
Your customers have a right to expect reproduceable results under normal conditions, not to have things happen acceptably one day and cause administrative action against them the next.
Either it should be legal to repost transactions to the batch system every N seconds, in which case you:
- figure out an acceptable number of customers doing so simultaneously
- configure for that plus margin
- monitor for acceptable performance
- add hardware or change your policy (with sufficient notice) if performance degrades
...or:
- put limits on the behavior
- bar violators consistently whether or not it has any affect on performance
