Summary Report # Samples counter rolls over and causes miscalculated fields for long running tests with high throughput

**Ryan Whitworth** ([Bug 57679](https://bz.apache.org/bugzilla//show_bug.cgi?id=57679&redirect=false)):
On the Summary Report listener page the # Samples field can roll over to zero during a long running test. This roll over then affects the values in the other fields that are calculated based on running time and the # of samples (average, throughput, KB/sec, etc).

This roll over happened during a 3 day test (over a weekend) when the throughput was roughly 35,000 connections a second. It looks as though the # Samples field is kept as a 32-bit number. Could this be increased to a 64-bit number so that longer tests could be run without fear of the averages becoming inaccurate?

OS: All

The issue concerns the Summary Report listener and its # Samples field; start by locating that listener and tracing where the sample count and average, throughput, and KB/sec values are calculated. Confirm the counter width and add or update focused coverage for rollover-safe calculations; done means long-running, high-throughput counts no longer wrap and the derived fields remain accurate.

Written by the indexing model from the issue text.