The Verification Test Is the Only Thing That Closes a Repair (§ 84.106(e))

There is a sentence contractors say on the phone that is not true, and it costs them files.

"We fixed it, so we're good."

You are not good. Under 40 CFR § 84.106, fixing the leak is the middle of the process, not the end of it. The thing that closes a repair — the thing an inspector looks for, the thing that lets the clock stop, the thing that unlocks three separate downstream benefits you are otherwise not entitled to — is a verification test. Two of them, actually.

Twenty-five posts on this site mention verification tests in passing. None of them has ever stopped to explain what they are. This one does.

What § 84.106(e) Actually Says

The requirement is one sentence long, and the word doing the work is both.

Read it carefully and two things fall out that shops routinely get wrong.

"Both." One test is not compliance. A tech who pressurizes the system, sprays the joint, sees no bubbles, and writes "leak repaired, verified" on the ticket has performed one test — the initial one — and left the file half-finished. Paragraph (d)(2) closes the door on any ambiguity: "Repairs of leaks must be documented by both an initial and a follow-up verification test or tests."

"On each leak." Not per appliance, not per service call. If the inspection found three leaks and you repaired three leaks, the verification obligation runs to each of them. A rack with a leaking flare, a leaking Schrader, and a weeping brazed joint is three repairs and three verification obligations, even though it is one truck roll and one invoice.

Test One: the Initial Verification Test

The initial test answers a narrow question: did the physical fix hold?

Notice the clock. The initial test is not on its own timer — it runs on the same 30-day clock as the repair itself, anchored to the exceedance, not to the day you finished wrenching. This is why a repair completed on day 29 is a bad repair schedule even when it's a good repair: you have one day left to test it, and if the test fails you have one day to redo the work and retest. The 30-day repair clock and the initial-verification clock are the same clock.

Then the rule gets specific about when in the job the test happens, and this is the part most often violated by accident:

• (e)(1)(i) — For repairs that don't require opening or evacuating the appliance: test after the repairs conclude and before any additional refrigerant is added.
• (e)(1)(ii) — For repairs that require evacuation: test before adding any refrigerant back.

Both branches say the same thing in different circumstances. Do not charge the system first. The temptation is obvious — the customer wants cooling, the tech wants to hear it run — but a test performed after the recharge cannot distinguish "the repair held" from "the repair leaks slowly and there's now enough refrigerant in it to mask that for a week." The regulation is not being fussy. It is closing the one loophole that would make the test meaningless.

And if the test fails, nothing bad has happened yet:

A failed initial verification test is not a violation. It is the system working. What is a violation is arriving at day 31 with an appliance still above the threshold and no filed extension explaining why. Techs who believe a failed test is a black mark tend to hedge the write-up, and a hedged write-up is worse than a failed test in a file.

Test Two: the Follow-Up Verification Test

The follow-up test answers a different, harder question: does the fix survive the machine actually running?

A brazed joint that holds under a nitrogen standing-pressure test at ambient temperature is not the same joint at operating pressure, with compressor vibration, thermal cycling, and a suction line that moves. The follow-up test exists because plenty of repairs pass the first test and fail the second one.

Three practical readings:

The 10-day window has two possible anchors, and they are not interchangeable. If the repair didn't require evacuation, the clock starts at the successful initial test. If the appliance was evacuated, the clock starts when it reaches normal operating characteristics and conditions — which is usually later. Get this backwards on a seasonal unit and you either test far too early to learn anything or you blow a deadline you thought you had. (The repair deadline calculator projects the follow-up date from the initial-test anchor — the conservative one — alongside the rest of the clocks.)

"Normal operating characteristics and conditions" is the whole point. § 84.102 defines the follow-up test as checking the repairs "after a successful initial verification test and after the appliance has returned to normal operating characteristics and conditions to verify that the repairs were successful." A follow-up test run on a system sitting idle in a mechanical room is not a follow-up test. It's a second initial test wearing the wrong label.

There is a genuine safety carve-out — and it's narrow. Where it's unsafe or impossible to test at operating conditions, the test moves to before the system returns to normal conditions, "where practicable." That's an accommodation for the ammonia room and the energized 480V section, not a general-purpose excuse. If you use it, the reason belongs in the record.

Same forgiveness applies here: (e)(2)(i) permits as many additional repairs and verification tests as needed to bring the appliance below the leak rate within the applicable time period.

Why the Follow-Up Date Is the Most Valuable Date in the File

Here is the part that turns a paperwork chore into something worth doing carefully. The date of the successful follow-up verification test is load-bearing in three separate places in Subpart C — and every one of them works in the contractor's favor.

1. It restarts the rolling-average window. The § 84.102 rolling-average method sums refrigerant added "over the previous 365-day period (or over the period that has passed since the last successful follow-up verification test showing all identified leaks in the appliance were repaired, if that period is less than one year)." Without a recorded follow-up test, a unit that leaked badly in March keeps carrying that March addition in its leak rate until the following March. With one, the window resets at the test — and the appliance stops looking non-compliant the day it stopped being non-compliant. If you are still calculating leak rates on a flat 365-day window after a verified repair, you are reporting a worse number than you are legally required to report.

2. It starts the inspection schedule clock. The § 84.106(g)(1) inspection frequencies — quarterly for commercial and industrial process refrigeration at 500+ lbs, annual for everything else covered — all run "after the date of a successful follow-up verification test," and they run until the appliance can demonstrate four clean quarters (or one clean year). The follow-up date is the start line for the post-exceedance inspection cadence. No recorded follow-up date, no defensible start line.

3. It starts the 12-month success presumption. This is the quiet gift in the rule:

A presumption is an evidentiary posture. Twelve clean months after a recorded follow-up test and the burden shifts: the repair is presumed successful rather than something you have to affirmatively prove years later to an inspector who was not there. That presumption is purchased with one date and one result written down at the time.

The follow-up verification test is the cheapest thing in Subpart C. Ten minutes with a detector buys a window reset, an inspection start line, and a legal presumption — and skipping it costs all three at once.

Who Is Allowed to Perform It

This question comes up constantly and the honest answer has a seam in it.

Paragraph (e) assigns the duty to the owner or operator: "The owner or operator must conduct both initial and follow-up verification tests." It does not, in its own text, repeat the certified-technician requirement that appears elsewhere in the same section. By contrast, § 84.106(d)(1) requires that "a certified technician must conduct a leak inspection… to identify the location of leaks," and § 84.106(g)(2) requires that "leak inspections must be conducted by a certified technician using method(s) determined by the certified technician to be appropriate."

So the leak inspection has an explicit certification requirement written into it. The verification tests, as drafted, do not.

That is a textual observation, not permission to hand a detector to an untrained helper. Two things constrain it in practice. First, verification testing that involves accessing the refrigerant circuit — pressure testing, evacuation, anything that could release refrigerant — pulls in the Section 608 technician certification requirements at 40 CFR § 82.161 on their own terms, independent of § 84.106(e). Second, a verification test performed by someone with no certification is a weak record. When the file is read by someone who was not there, "electronic detector, no indication at repaired joint, tested at operating conditions — J. Martinez, EPA 608 Universal" is evidence. An unsigned line that says "checked, ok" is not.

Use a certified technician. Record who it was. The regulation's silence is not an invitation.

The Record § 84.106(l)(7) Requires

The recordkeeping obligation is one line, and it is broader than most people assume:

All of them — including the failed ones. A repair that took three attempts produces three initial tests and their results, not just the one that passed. The failures are not embarrassing; they are the evidence that you were working the problem inside the window. A file showing three dated attempts and a pass on day 26 tells a coherent story. A file showing one test on day 26 with no history invites the question of what you were doing for the first 25 days.

The paragraph (l) chapeau sets retention: "All records identified in this paragraph must be kept for at least three years in electronic or paper format, unless otherwise specified." Verification test records fall under that default — three years, and electronic is explicitly fine.

What a defensible verification record contains, in practice:

Where This Goes Wrong in Real Files

Four failure patterns account for nearly everything we see.

The one-test close. Initial test performed and recorded, follow-up never happens, repair marked complete. The appliance is out of compliance with (e) and (d)(2), the rolling window never resets, and the inspection schedule has no anchor. This is the single most common gap, and it happens because the truck left the site.

The post-charge test. Correct work, correct detector, wrong order. Fails (e)(1)(i)–(ii). Costs nothing to fix and nearly always survives unnoticed until someone reads the ticket sequence.

The idle follow-up. Follow-up test performed on a system that isn't running, usually because the tech was on site for something else and it was convenient. Doesn't meet the (e)(2) / § 84.102 operating-conditions standard.

The undated pass. "Verified — J.M." with no date, or a date that's the invoice date rather than the test date. (l)(7) asks for dates and results. An undated pass cannot anchor a 10-day window, a window reset, an inspection schedule, or a 12-month presumption — which means it silently forfeits all four.

There is also a timing subtlety worth knowing: if the appliance is mothballed, § 84.106(d)(3) temporarily suspends the paragraph (d) through (f) time frames, and they resume on the day refrigerant is next added. The verification clocks pause with everything else — but the pause has to be documented too.

The Shape of It

Strip out the citations and Subpart C's repair sequence is four moves:

1. A certified technician finds the leaks.
2. You repair them inside 30 days of the exceedance.
3. You prove it twice — once before the refrigerant goes back in, once after the machine is running normally.
4. You write down both dates and both results, and keep them three years.

Step 3 is the one shops skip half of, and step 4 is the one that converts the work into a defense. The follow-up verification test is ten minutes of a technician's day. It is also the date the rolling window restarts from, the date the inspection schedule counts from, and the date the twelve-month presumption of success begins. No other single entry in the appliance file does that much work.

Fixing the leak was never what closed the repair. The test is.