The Parts Condemned During Someone Else's Recovery

Why this matters

There is a specific kind of already-replaced part that nobody handles well: the one swapped in the first fifteen minutes after power came back. The equipment was doing something normal and temporary, somebody read it as dead, a part went in, the recovery window expired on its own schedule, and the system started working. Everyone in the story now believes the part fixed it. Nobody is lying and nobody is right.

You inherit this on the callback, and it is harder than an ordinary parts-cannon job, because two things happened at once: a component you did not install is now in the circuit, and the condition that produced the original symptom evaporated before anyone measured it. Treating this like a normal second opinion, where you re-run the original diagnosis, fails immediately - the original diagnosis is not re-runnable.

Check the installation before you check the theory

Whoever put that part in may not have terminated it the way you would. Before energizing anything to test a theory, de-energize at the disconnect, lock and tag it, and verify dead before you touch terminals, which is the duty at 29 CFR 1910.333(b)(2) for work on electric circuit parts, with 29 CFR 1926.417 as the construction counterpart; prove dead using the live-dead-live sequence in NFPA 70E-2021, 120.5. Inspect the new part's terminations, polarity, orientation and mounting with the power off, and pull on each conductor at its landing. If the replacement touched a gas train, a flue, or a combustion control, the appliance does not run for a test until the connections and the venting have been proven leak-free and intact. If it sits on a compressor, an accumulator or any pressurized or spring-loaded assembly, isolate and bleed or block the stored energy under 29 CFR 1910.147 before service.

A badly landed new part is the second most common finding on these calls, behind the part having been unnecessary in the first place, and it is the one that hurts somebody.

The gate: when was the decision made, relative to the recovery window

One question sorts these calls, and it is chronological rather than technical.

Reconstruct the recovery window for that equipment - two times the longest published delay in its own service literature, floor of 15 minutes, counted from the moment supply became continuously present - and ask whether the condemnation decision was made at or inside that window. The unit of analysis is per piece of equipment, not per site: two units on one property can have different windows off the same outage.

• Decision made inside the window: the condemnation was made during the only period when a healthy system reliably looks dead. It carries close to zero diagnostic weight. Treat the new part as a variable AND treat the original complaint as unproven, both.
• Decision made outside the window, on repeating evidence: the condemnation had something behind it. The new part is a partly-validated constant, and your question shifts from "was this needed" to "why did the symptom survive it."

The reason chronology beats technique here: you cannot test the old part's condition on the day it was pulled, but you can almost always establish when it was pulled. Time is the one piece of evidence a replacement does not destroy.

What a replacement destroys, and what survives it

Build your approach around the second column. The first column is gone and no amount of care recovers it.

Ask for the old part before you ask anything else, and ask in a way that gets a real answer: "Is the part that came out still here, or in the truck, or did it go in the bin?" A component that is merely out of the circuit still tells you its rating, whether it was swollen, vented, discolored, burnt at a terminal, or physically undamaged and clean. A clean, correctly-rated old part is strong evidence that it was never the problem.

Promoting a new part from variable to constant

A part you did not install is not a known-good just because it is new. New parts arrive defective, get damaged in handling, and get installed wrong. State the rule per part:

A newly installed part becomes a constant only if at least one of these is true: you verified it in circuit under load yourself, or you installed it yourself and made the terminations, or its failure mode is directly observable and you observed the part not exhibiting it. The Boolean is OR, one is enough. Fail all three and it stays a variable on your suspect list, and it stays there even when it is the newest thing in the box.

The practical consequence: on these calls, promoting the new part is usually your first move rather than a formality, because every subsequent measurement is ambiguous while it is still a variable. It is also fast, since verifying a component in circuit under load is generally a single reading against a marked rating.

Running the gate against two calls

Call one: the swap that happened at minute twelve. An outage of about 25 minutes, three interruptions counting two fast flickers before the sustained loss. The literature for this equipment publishes one relevant delay, a 5-minute hold-off, so the window is two times 5, which is under the 15-minute floor, so the window is 15 minutes from continuous restoration. The customer's brother-in-law swapped a start component about 12 minutes after power returned, and the system started working shortly after. That decision sits inside the 15-minute window, with 3 minutes of window still to run.

So: the new part is a variable and the original complaint is unproven. The tech verifies the new component in circuit under load, reads it within tolerance of its own marking, and promotes it to a constant. Then he asks for the old part. It is in the truck, clean, correctly rated for the application, not swollen, not vented, no terminal discoloration. Two independent findings now point the same way: nothing was demonstrably wrong, and the symptom that triggered the swap was a hold-off that had roughly a fifth of its window left to run when the part came out.

The correct output is not a repair. It is a documented no-fault-found with a named tripwire: if the equipment fails to produce output more than 15 minutes after any future power event, or trips any protective device at any time, that is a real fault and it gets diagnosed then. The tech also warns the customer plainly that the swap has not been validated as the fix, because the alternative is a household that now believes this part fails routinely and buys another one next season.

Call two: the swap that happened three days later. Same kind of outage. This time the customer lived with it, the equipment tripped a protective device six times across the following two days, and a part went in on day three. That decision sits well outside any recovery window, on a repeating protective trip, which is on nobody's list of recovery behaviors. The condemnation was defensible.

Here the tech promotes the new part the same way, verifies it under load, and then asks the different question: the trips did not stop, they changed. They now happen only when a second piece of equipment runs at the same time, which was never true before. That is not a failed diagnosis, it is a fault that moved, and the investigation goes to the shared supply rather than back to the component. Same gate, same two minutes of work, opposite conclusions and opposite next steps.

Saying it to the customer without indicting anyone

The customer paid for that part and often paid a relative or a handyman to install it. Telling them it was unnecessary lands as an accusation unless you frame it as timing rather than competence, and an accused customer stops giving you information.

What works: "Equipment holds itself off for the first several minutes after power comes back, by design, so it does not restart against pressure. Yours had about 15 minutes of that. The part went in at minute twelve, so we cannot tell from this whether it was ever bad. The part in there now tests good, so we are fine either way, and here is the specific thing to watch for that would mean there is a real problem."

That version gives them the mechanism, keeps the old part's status honestly open, and hands them the tripwire. It also protects you, because when the equipment runs fine for a year you were never the one who claimed the swap fixed it.

How to verify you handled it correctly

• You physically inspected the new part's terminations with power off. Not glanced at, not assumed because it works. This is the finding that hurts somebody if you skip it.
• You either have the old part in your hand or you have written down that it was discarded. "Unavailable" is a legitimate record. Silence about it is not, because the next tech will assume you looked.
• Your write-up names which case you were in. Inside the window or outside it, with the window arithmetic and the clock start shown. A note that says only "customer replaced part, tests good" tells the next person nothing and forces them to re-derive the timeline from a customer whose memory has now had three weeks to soften.
• If you found no fault, you left a tripwire the customer can actually observe. A condition with a time on it, and a protective-device trip. Not "call if it acts up."

References

• 29 CFR 1910.333(b)(2), OSHA electrical safe work practices, for de-energizing and locking or tagging before work on electric circuit parts; 29 CFR 1926.417 is the construction counterpart for electrical lockout and tagging
• NFPA 70E-2021, 120.5, the live-dead-live process for establishing an electrically safe work condition
• 29 CFR 1910.147, OSHA control of hazardous energy, for isolating stored energy before service on pressurized or spring-loaded assemblies
• Manufacturer service literature for published restart delays and hold-off intervals used to reconstruct the recovery window
• See related: What a System Does While It Is Recovering From an Outage; Reading the Evidence of a Previous Repair Attempt; The Parts Cannon: Why Guess and Swap Costs More Than It Saves