Monitoring a volume bot run
The watch loop for a run that is already live: what each check is meant to answer, where the answer comes from, how often to look, and the run journal that makes the next hour readable.
01OpenThe Live Window: The Hour a Run Is Open
A run has three phases and only one of them is covered here. Configuration happens before, review happens after, and in between there is a window where the software is signing and submitting and you are the only person watching. This section is about that window: what to look at, how often, what to do about what you see, and how to end it properly.
The window has a shape. The first minutes establish a baseline you will compare everything against. The middle stretch is where discipline decays and unnecessary adjustments creep in. The last stretch is where operators relax too early and miss the transactions that have not settled yet. Each of the three pages below owns one of those problems.
phase From the first submitted transaction to the last settled one, including the part after you press stop
inputs Whatever your console reports, whatever the chain confirms, and whatever you wrote down before opening
outputs A decision every time you look, where taking no action is one of the available decisions
ends when The plan is complete, a stop condition fires, or you decide the run has stopped being worth its cost
3 pieces in this section
The hour the run is actually open. Watch cadence, what a check is supposed to answer, which adjustments are worth making mid-flight, and how a run is closed out once the last transaction has settled.
When to intervene and when not to
An escalation ladder for mid run adjustments: observe, adjust, pause, stop, abort. The two-reading rule, the reversibility test, and what over-intervening does to the record you keep.
02OpenClosing a run and reconciling
Stopping is not closing. The quiesce sequence, the four ledgers that have to agree, the costs that only appear at the end, and the run report that sets the thresholds for next time.
03OpenHow this section is used
Read the watch loop first, because it defines the cadence everything else assumes. The intervention ladder is what you reach for when a reading is uncomfortable, and the closing procedure is the one people skip, which is why the numbers so often fail to add up a day later.
Signals What a live reading is made of and what it is allowed to trigger. Signal families, sampling windows, the thresholds you set before the run starts, and the conditions that end it without a debate. Questions Twelve short answers to the questions that arrive while a run is open, from handover etiquette to telling a finished run from a stalled one.