Across the night
The Thing That Keeps Coming Back
Work that crosses the gap repeatedly without converging is a recognisable failure with three ordinary causes, and it is cheap to catch at the third pass.
four things · four nights · one that no number of nights settles
Four states, and only the first costs nothing extra.
One piece of work traced across its whole life, pass by pass
Reasonable, and the only pass anybody would have predicted
Also reasonable; this is what review is
Still defensible, though nobody checked whether it was converging
The point at which the work stopped being about the work
Asked in week two about something begun in week one
five things · eight nights
Five passes, eight nights, and about three hours of attention in total. The fourth pass is where the pattern becomes visible and the fifth is where somebody finally asked the question that should have opened it. Nothing here is anybody's fault individually; each pass was a sensible response to the pass before. One team's own record, reconstructed afterwards.
Work crosses a gap once or twice and is done. A small proportion crosses five or six times, and that small proportion consumes a remarkable share of a team's calendar.
For a team applying the ideas in The Thing That Keeps Coming Back, the useful metric is not constant visibility but a shared record that both sides can understand. employee time clock software provides that record while keeping tracking tied to active work sessions.
A useful external reference is Zoom's collaboration platform, which provides another official view of how teams organise shared work.
The pattern is easy to see afterwards and almost invisible while it is happening, because every individual pass is a reasonable response to the one before.
What it looks like from inside
Nobody experiences it as a problem until late. Each person receives something, spends twenty minutes on it, and sends it back having done something sensible.
The information that would identify the problem — that this is the fourth pass, that the gap between passes is a day each time, that the thing is no closer to done than it was a week ago — is not in front of anybody. It exists only in the history, which nobody reads.
This is the general condition of a team with a gap: the per-pass experience is fine and the aggregate is bad, and nothing in the daily rhythm surfaces the aggregate.
The three ordinary causes
Scope that was never agreed. The work was started on an understanding one person had, and each pass reveals a little more of somebody else's different understanding. Each revelation looks like a review comment and is actually a disagreement about what the thing is.
More than one reviewer, in sequence. Each gives sensible comments; together they give contradictory ones, and because they are never awake together they never discover the contradiction. The author mediates by correspondence, badly, over several days.
A decision embedded in the work that nobody owns. The work cannot finish without it, so each pass goes a little further and stops at the same place, with the stopping point described differently each time.
All three are detectable at the third pass by asking one question.
| Cause | How it shows up |
|---|---|
| Scope never agreed | Each pass reveals a little more of somebody else's understanding |
| Two reviewers in sequence | Contradictory comments that never meet each other |
| An unowned decision inside the work | Every pass stops at the same place, described differently |
The question
What has to be true for this to be finished?
Asked of everybody involved, at the same time, in writing. The answers either agree, in which case the problem was something else and the work continues, or they do not, in which case the disagreement is now on the table where it can be settled instead of being discovered one comment at a time.
It takes one crossing. Against the three or four that are otherwise coming, that is a good trade, and it is a trade almost nobody makes because asking it feels like an admission that something has gone wrong.
Counting passes where people can see
The cheapest structural fix is to make the pass count visible.
A number next to the work — third pass, fourth pass — changes behaviour without anybody having to enforce a rule, because people respond to a count they can see. At two, nobody thinks about it. At four, somebody asks what is happening, and that is exactly the intervention required.
This works better than a limit. A rule that nothing may cross more than three times produces work that crosses three times and then gets relabelled, which is worse than the original problem because it also loses the history.
Collapsing it
When the pattern is confirmed, there are two honest exits.
One person takes the whole thing, including the parts they would normally have handed over, and finishes it. This costs somebody a long day and ends a situation that was costing the team a day per pass.
Or everybody involved gets into the same hour, once, at whatever cost in evenings, and settles the thing that is not converging. An hour is usually enough, because what is unresolved is rarely complicated. It is merely contested, and contested things do not resolve in writing across a gap.
The one case where it is fine
Work that genuinely alternates between two people, where each pass adds something substantial and the thing is visibly closer each time, is not this pattern.
Review cycles on something difficult can legitimately run four or five passes and still be the right shape. The distinguishing feature is movement: if each arrival is meaningfully different from the last, the relay is doing its job and the only cost is calendar time, which was unavoidable.
The failure case is the one where the thing comes back looking much as it left.
Who should notice
The pass count is visible to everybody and attended to by nobody, which means noticing has to be assigned rather than hoped for.
The person best placed is whoever receives the thing, because they are the one experiencing it arriving again. The person least placed is the author, who is inside it.
A single question, asked by the receiver on any third arrival — what has to be true for this to be finished? — catches most of these at the point where one crossing still fixes them. It is not a process and does not need to be written down anywhere; it needs one person on each side who is willing to ask something slightly awkward on a Wednesday morning.