Diagnosing agent session problems with the Latency Gaps Report

When an agent says their screen "froze for a bit" or "logged itself out" and you can't reproduce it, you need evidence rather than a guess. The Latency Gaps Report gives you that. It surfaces the moments where the agent screen stopped reporting in while the agent was still supposed to be active — and those silent stretches are usually the root of the session problem you're chasing.

What a latency gap actually is

The agent screen checks in with the server on a regular heartbeat and records the round-trip Latency of each check. A latency gap is a missing segment of those log entries during a window when the agent was supposed to be logged in. In other words, the screen went quiet — it stopped phoning home — even though the Agent session was still open. The Latency Gaps Report lists each of those gaps, and it can chart them for all agents across a single day so clusters jump out.

Why gaps explain session problems

When the agent screen stops checking in, several things can break at once. The server may decide the agent went away and tear down the Session ID, leaving the agent's Status (lead status) stuck or kicking them out. Calls that should route to that agent may skip them. The agent sees a frozen screen, a spinning request, or a sudden logout — and a gap in the log lines up exactly with the moment they describe.

Reading a gap from cause to symptom

flowchart TD
  A[Agent screen heartbeat stops] --> B[Latency gap recorded]
  B --> C{How long was the gap?}
  C -->|Brief blip| D[Transient network hiccup]
  C -->|Sustained| E[Server may drop the session]
  E --> F[Agent status stuck or logged out]
  D --> G[Screen catches up and resumes]
  F --> H[Correlate with the agent timeline]

Start with the agent and the time the user complained about, then look for a gap in that window. A short blip is usually a transient network hiccup that the screen recovered from. A long, sustained gap is the kind that lets the server drop the session — and that's the one worth investigating. Cross-check the gap against the same agent's normal Latency readings around it to see whether the connection was already getting slow before it went silent.

Where the gap usually comes from

Most gaps trace to one of three places: the agent's own network or workstation, the path between them and the server, or load on the server itself. If gaps cluster around one agent only, look at their machine and connection. If gaps cluster around one time of day across many agents, suspect a shared bottleneck — a saturated uplink or a server under heavy load. The chart in the report is the fastest way to tell those two patterns apart.

For the field-by-field walkthrough of this report, read how to read the Latency Gaps Report. When the gaps are real but you need to understand the slow connection behind them, diagnosing high agent latency is the next stop, and the VICIdial troubleshooting playbook ties it all together.

A loaded, under-resourced server produces gaps no agent-side fix can solve. VICIfast runs your dialer on a dedicated, hardened box that's live in under 40 seconds, so server load is one less thing creating gaps. See our plans and pricing.