Well, "reasonable" is sometimes subjective. Just doing some basic maths here -- 82,843,934 events / 2,413 secs = 34,332 events per second scan rate. If you divide that by the number of indexers (assuming the data is perfectly distributed, which may not be true) that is 11,444 events per second per indexer.
Another assumption that each event is 1000 bytes (which may not be true) puts your throughput around 11 MBytes/sec - which is low relative to the basic throughput of a modern disk subsystem. You do have an appropriate disk subsystem attached, right? And these are physical machines, or VMs?
This also includes overhead from search-head to indexer coordination, CPU-time cost of doing field extraction, and a few other things. You really don't have the information to see where all the time was spent. There's a search job inspector tool that can help. Perhaps you can update with data from it?
But, I think there is a bit of misconception here. A search to "count ALL the things!" is not really a objective test of search performance. You need to search for something other than "everything". A highly dense search (where the number of events returned is a large fraction of the total number of events in the system) will usually be slower than a relatively sparse one.
