What this is
On August 3 this series published The Form 1 Backlog Broke, which reported that Form 1 Individual waits had reversed after five months of climbing, hedged the finding hard, and promised a full reassessment at the end of August. On August 18 it published an interim check because one question could not wait: whether Form 1 Trust was following. That piece answered it and deferred everything else to this one.
This is that reassessment. It covers 4,006 approvals reported by NFA filers, snapshotted August 31, 2026. Community data, not official ATF records, with one exception that is labelled where it appears. The method note at the end covers what was excluded and why.
Three questions were left open. All three now have answers, and one of them is the answer we spent two months saying we could not give.
The three open questions
Does Form 1 Individual settle in the low twenties, or drift back up? Neither, again. The lane kept falling. Over the fourteen complete days ending August 30 it ran a median of 3 days on 50 approvals, with a 95% confidence interval of 2 to 3 days. Its 90th percentile is 4 days, meaning nine in ten reported approvals in that window cleared inside four days.
Does Form 1 Trust keep going? Yes, and it moved further in the second half of August than the Individual lane did in the same period. Splitting the month, Form 1 Trust ran 59 days over August 1 to 15 and 16 days over August 16 to 30, on 62 and 38 approvals, with confidence intervals that do not overlap.
Does Form 4 Trust keep going? This is the one the August 18 edition explicitly declined to call. Its weekly series looked suggestive, the formal two-window test said too close to call, and the piece said that a suggestive series plus a failing test is exactly where this series has previously talked itself into a finding. Two weeks of additional records settle it. The two-window comparison the site runs on the Queue State dashboard now puts Form 4 Trust at 31 days over the thirty days to July 28 and 24 days over the thirty after, on 83 and 97 approvals, which clears the bar. Waiting was the right call and the answer was yes.
All Four Lanes, Weekly
Median approval wait (days) by the week ATF cleared the filing, complete weeks only. The two Form 1 lanes fall from the top of the chart to the bottom. What is left at the end is not a gap between Form 1 and Form 4.
Complete approval weeks only. Sample counts vary by lane and week; a single thin week is not a level.
The comparison this series has been getting wrong all year
Every edition since June has measured the fall against the peak. That answers whether things are better than they were in July, which was never really the question anyone was asking.
January is the right comparison. It is the earliest month this dataset covers with a usable sample, it is the last month before the climb started, and it has one property that makes it a fair benchmark rather than just an old number: 547 of its 548 approvals were filings placed that same month. January was a queue working current filings, which is what late August is. The two months are the same kind of thing measured twice.
The Whole Year, Month by Month
Median approval wait (days) by approval month. January is the last month before the 2026 Form 1 slowdown began, so it is the level the year has to be read against rather than the spring peak.
August covers approvals through 2026-08-31 and is a partial month. A month-long median averages over a lane that moved inside the month.
Against that baseline, over the fourteen days ending August 30:
- –Form 1 Individual ran 13 days in January and 3 days now, on 142 and 50 approvals. The intervals are 12 to 14 and 2 to 3, so they do not overlap. This lane is faster than it was before the slowdown began.
- –Form 4 Individual ran 11 days in January and 4 days now, on 195 and 52. Also separated, also faster.
- –Form 4 Trust ran 10 days in January and 16 days now, on 124 and 43. The intervals are 9 to 13 and 13 to 24, which touch. We cannot say this lane is slower than its January self and we cannot say it has fully returned.
- –Form 1 Trust ran 12 days in January and 13 days now, on 75 and 31. Intervals of 9 to 14 and 6 to 16, overlapping comfortably. Back to where it started, as far as this data can tell.
The 2026 Form 1 slowdown, which took the Individual lane from 13 days in January to 69 in June, is over. Not easing. Over, and past its starting point on both individual lanes.
What replaced the split
From February onward this queue had a shape, and the shape was the form. Form 4 Individual sat near a week all spring while Form 1 Individual climbed past sixty days. June's edition called it a two-queue regime and July's called it a freeze. Both were describing the same thing, which was that the form you filed determined roughly how long you waited.
That is gone. Here is where the four lanes actually stand.
Where the Four Lanes Stand
Median approval wait (days) by lane over the last 14 complete days, with a 95% confidence interval on each median. The two individual lanes overlap each other and the two trust lanes overlap each other, so on this evidence the data separates these four lanes into two groups rather than four.
Read the confidence intervals rather than the bars. Form 1 Individual is 2 to 3 days and Form 4 Individual is 3 to 5, so the data cannot separate them. Form 1 Trust is 6 to 16 and Form 4 Trust is 13 to 24, which also overlap. Four lanes, and the evidence supports sorting them into two groups.
The grouping is by owner type. An individual filer waits a few days on either form. A trust waits something in the region of two weeks on either form. The penalty for filing through a trust is about twelve days on Form 4 and thirteen on Form 1, and in both cases the individual and trust intervals are cleanly separated, so the gap itself is real even where its exact size is not pinned down.
This is worth stating carefully because it inverts a thing this site has published repeatedly. The July edition ran a whole section on the Form 4 trust penalty having doubled and locked near 21 days, and treated it as the stable structural feature of an otherwise chaotic year. It is now the only structural feature left, and it is smaller than it was.
What each lane is working
The strongest evidence for what happened is not a wait time. It is which filings are being approved.
How Much Old Work Each Lane Is Still Clearing
Share of each week's approvals that were filed in an earlier calendar month. A lane running at 100% is working a backlog. A lane at 0% has nothing older than the current month left to approve.
Complete approval weeks only.
A lane clearing a backlog approves filings placed in earlier months. A lane that has run out approves the month it is in. Through mid-July all four lines sit high. In the week of August 24, Form 1 Individual approved 23 filings and every one was placed in August. Form 1 Trust approved five, all placed in August. Form 4 Individual approved 22, all placed in August.
Form 4 Trust is the exception and the only lane with old work left. It approved 24 filings that week, seven of them placed before August. Two weeks earlier that share was 76%. It is draining rather than drained, which is consistent with it being the last lane to turn and with its median being the highest of the four.
The distributions say the same thing in a different way. Over August 16 to 30, 98% of Form 1 Individual approvals and 92% of Form 4 Individual approvals cleared within a week, and the slowest single approval in either lane waited 25 days. Form 4 Trust put 15% inside a week and its slowest waited 41 days. Form 1 Trust is the interesting one: 32% cleared within a week and another 42% took between fifteen and thirty days, which is a lane caught partway through its own drain rather than one sitting at a level.
Why, which we can now answer
Every edition since July has carried the same disclaimer. We see approvals and never applications, so a capacity increase and a backlog running dry look identical from where we sit. The August 18 edition said it again.
We can retire it. ATF publishes year-to-date application counts, this site banked a reading of them on June 1 and another on August 5, and two readings of the intake side separate what one reading cannot.
Year to date on June 1, ATF had received 161,478 Form 1 applications and finalized 108,608, leaving 52,870 outstanding. On August 5 it had received 202,599 and finalized 183,948, leaving 18,651.
The outstanding Form 1 queue fell 64.7% between those readings. Form 4's fell 13.7% over the same period. In the roughly two months between them Form 1 finalized 1.83 applications for every one it received, against Form 4's 1.03.
Both mechanisms moved, and neither accounts for the drawdown by itself. Form 1 intake fell by roughly 41%, from about 32,300 a month over the first five months to about 19,000 a month between the readings. Form 1 throughput rose by roughly 60% over the same comparison, from about 21,700 finalized a month to about 34,700. Treat those as magnitudes rather than measurements, because the two "as of" dates are approximate and the windows are unequal. The direction of both is not in doubt.
So the answer to the question this series has been refusing for two months is that it was both, in roughly similar measure, and that is why the fall was as fast as it was. Fewer applications arriving and more being worked is a queue emptying from two ends.
One further figure from that reading predicted what we now observe. In July, ATF reported 62,916 actionable NFA applications against 254,639 finalized, which is about 7.4 days of work at July throughput. A queue holding a week of work is a queue at working capacity, and single-digit medians on both individual lanes are what that looks like from the outside.
⚠️ "Finalized" is ATF's word and it does not mean approved. Their footnote defines it as approved, disapproved, withdrawn, or returned without action. A drawdown computed from it is a drawdown of work, not of approvals. Our data cannot see disapprovals at all, so the two sources are measuring adjacent things rather than the same thing.
What a Form 1 is for
The August 3 edition listed three things it could not tell anyone. One was whether suppressor and short-barreled rifle Form 1s had moved together, because nothing in the data told them apart.
That changed quietly in the middle of July, when the intake form started recording what the filing is for. Coverage is now 60% of records ingested in August and effectively zero before July, so this can be asked about recent windows and nothing else.
Before using it, the obvious worry is that people who answer the question are not like people who skip it. Across all four lanes in August, the labelled and unlabelled records differ by at most one day in median. Whatever else it is, the field is not selecting for fast or slow filings.
Among Form 1 Individual approvals over the fourteen days ending August 30, short-barreled rifles ran a median of 3 days on 23 approvals and suppressors 2 days on 15, with intervals of 2 to 4 and 2 to 3. They moved together. On the whole of August the same split reads 6 days and 3, but that window spans the fall itself and the two item types are not evenly spread across it, so we are not leaning on it.
⛔ Do not read an SBR penalty out of aggregate Form 1 numbers. Trust filings in our August sample are about 91% short-barreled rifles and individual filings about 70%, and trusts are slower. Pooling the two owner types makes SBRs look slow when what is actually slow is trusts.
For Form 4, where suppressors are most of the volume, August suppressor approvals ran a median of 3 days on 45 individual filings and 22 days on 54 trust filings. The same two-tier structure, in the lane that most suppressor buyers actually use.
The checks
Slower reporting cannot be manufacturing this. Recomputing the August halves using only records reported within three days of approval holds the reporting window constant across both. All four lanes give the same answer either way: Form 4 Individual 5 to 4, Form 4 Trust 28 to 16, Form 1 Individual 7 to 3, Form 1 Trust 59 to 16.
The Form 1 Trust Series, Measured Two Ways
Form 1 Trust weekly median computed on all records and on only those reported within three days of approval. The second series holds the reporting window constant, so slower reporting cannot manufacture a trend. Where the two diverge, that week's figure rests on early reports and should be read as provisional.
No single state is carrying it. Over August 16 to 30, dropping the most common state from each lane moves its median by at most one day. Form 4 Trust's most common state is 26% of that lane's records and dropping it leaves the median unchanged at 16 days. This matters because a state cluster is exactly what distorted a week of June, and the check that caught it then is the check being run here.
Filings still in the queue cannot be hiding in these numbers. Counting only approvals that waited 30 days or less removes the question entirely, because a filing that cleared in under a month cannot be one still sitting there. Form 1 Trust produced no such approvals in most weeks through mid-July. In the week of August 17 it produced 24 out of 26.
Form 1 Approvals Clearing in 30 Days or Less
Count of each week's Form 1 approvals that waited 30 days or less, restricted to records reported within three days of approval so every week is measured through the same window. A 30-day wait cleared this week was filed a month ago, so this count cannot be distorted by filings still waiting.
One number in the charts should not be read. Form 1 Trust in the week of August 24 rests on five approvals, against 33 and 26 in the two weeks before it. Form 4 Trust reported 24 that same week, so this is specific to one lane rather than a general reporting gap, and the most likely explanation is that more of that week has not arrived yet. Whatever the reason, five records will not carry a level and we are not publishing that week's figure as one.
What we still cannot see
We observe a sample. These are approvals reported by filers in the NFA community, and the people who report are not a random draw from everyone who files. This binds harder now than it did in the spring, because a fast lane turns its reported sample over quickly and the composition can shift without the queue changing.
We cannot see denials. Nothing in this data distinguishes a queue that got faster from one that got faster partly by returning more applications without action. ATF's finalized counts include those and ours do not, which is one reason the two sources are not interchangeable.
We can only partly detect batch filings. Concentration across the recent weeks is unremarkable on both measures available. A single party filing across several states over several weeks would still leave no signature we could find.
The item type field is two months old. It cannot answer whether suppressor and SBR Form 1s diverged during the climb, only whether they differ now.
What happens next
The thing worth watching is no longer whether the numbers fall. It is whether a queue running at working capacity stays there once the reason it emptied stops applying.
Two of the three things that produced this were temporary in character. A backlog can only drain once. Intake down 41% is a change in filer behavior, and nothing about it is obviously permanent, particularly with the tax stamp still at zero. The throughput increase is the one that could persist, and we have no visibility into what produced it.
So the question for September is narrower than the one this series has been asking since February. Not whether Form 1 recovers, which it has, but whether the individual lanes hold a three to five day median through a month where they are working only current filings and have no backlog left to make progress against.
Form 4 Trust is the lane to watch for a different reason. It is the only one with old work left, it turned last, and it is the only lane whose January comparison still leaves room for it to be genuinely slower than it was.
Method and limits
Figures come from nfa_watch.approvals as of August 31, 2026: 4,104 records carrying both a filing date and an approval date. One record with an implausible wait was excluded. Ninety-five were excluded as duplicates, on the rule that a duplicate requires both an identical (form, owner type, state, filing date, approval date) tuple and an identical masked control number. Our control numbers are truncated for privacy, so matching on the mask alone would discard genuinely distinct filings. Two paper filings were excluded, leaving 4,006 records analysed. Paper volume is negligible in our sample and 1.51% of ATF's year-to-date total, so excluding it is defensible from both sides, but our sample under-represents paper by roughly thirty times and no claim about paper should ever be made from it.
Medians group by approval week or approval month, which is the output side of the pipe, meaning what actually got cleared in that period. That is a different view from filing-cohort curves and should not be read as one. Medians are rounded to whole days. Only complete weeks appear in the weekly charts; the partial week containing August 31 is excluded. August is a partial month in the monthly chart, covering approvals through August 30. Confidence intervals are exact order-statistic intervals computed from the observed records, and they describe sampling uncertainty only, not the gap between this sample and the queue as a whole.
Reporting lag in this snapshot runs 50% of records within one day of approval, 81% within three, and 94% within seven. Where a chart is labelled as using a constant reporting window, it includes only records reported within three days, so every week in it is measured through the same aperture.
The ATF figures in "Why, which we can now answer" are ATF's own published statistics, banked by this site on two dates because that page overwrites itself monthly and has no archive or API. They are not our data and are labelled as such wherever they appear.
Charts on this page are locked to the August 31, 2026 snapshot and will not move. The live versions of these studies, which update continuously, are in the Lab. Current medians across all four lanes are on the wait times page, and this edition is dated and is left as published.