Fails Warm, Runs Fine Cold (Decision Tree)

Why this matters

This is the fault that makes a customer sound unreliable and makes a tech look bad: it runs perfectly when you show up (usually because it just sat and cooled, or because you are testing it fresh), and the customer swears it dies every afternoon. The pattern is real, it is just invisible on arrival because you are looking at the wrong thermal state. This tree gets you to the right test window fast, without a full-day stakeout.

Start here: confirm it is genuinely warm-triggered

• It starts and runs fine from a cold or rested state, including your own arrival test.
• It fails, shuts down, or misbehaves only after a run of some length, or once ambient and operating conditions have pushed it genuinely warm.
• It recovers on its own once it sits and cools, without any part being changed.

If it also fails intermittently right from a cold start, this is not a purely warm-triggered pattern; see the cold-start tree instead, or suspect a marginal component that fails regardless of temperature.

Safety note: a unit that fails specifically once it gets hot is very often failing because it is genuinely overheating, and a safety device shutting it down is doing its job, not malfunctioning. If you find scorching, an unusually hot enclosure, or smoke, treat it as an overheating hazard and de-energize before you probe further. Do not force a restart on a unit that shut itself down hot until you know why.

Step 1: rule out genuine overheating first

Before suspecting a component fault, check whether the unit is actually running too hot for legitimate reasons: blocked airflow, fouled cooling surfaces, a failing fan, a mismatched load, or an overload condition. If a high-limit or thermal protector is tripping after a real heat buildup and resetting once cool, the fix is the heat source, not the limit switch. Replacing a limit that is correctly protecting an overheating unit is a mistake that leaves the real cause running.

Step 2: if the unit itself is not overheating, the fault is a thermal-sensitive component

Once you have ruled out genuine overheating, the failure is coming from a specific part whose electrical or mechanical behavior shifts once it reaches operating temperature: a marginal connection expanding open, a capacitor losing capacity as it warms, a semiconductor or control board failing conduction hot, a winding with a weak spot that bridges or opens under thermal expansion, or a mechanical clearance or bearing that closes up and binds once warm. Each of these has its own specific test approach and read, covered in full in the companion article: see Works Cold, Fails When Hot: A Thermal Fault Tree for the component-by-component branch walk (connections, capacitors, semiconductors, windings, and mechanical binding) once you have confirmed you are past the overheating check in step 1.

Step 3: catch it in the actual failure window

You cannot diagnose a warm fault on a unit you tested cold. The method that actually works:

1. Take baseline readings cold, on arrival, before it has run at all.
2. Let it run until it fails, or as close to the failure point as is safely possible.
3. The instant it fails, take the same readings again, before it has a chance to cool: voltage at the load, current draw, and the temperature and continuity of your suspect connections.
4. Compare hot readings to the cold baseline. The component that changed between the two is your fault, not whatever tested fine on a single cold check.

Carry an infrared thermometer to find the hot spot quickly and a clamp meter to read current without breaking the circuit. Patience is the actual skill here: staying through the failure rather than concluding "no fault found" from a clean cold test.

The recap

1. Confirm the pattern is genuinely warm-triggered, not present cold too.
2. Rule out real overheating first; do not replace a safety device that is correctly protecting the unit.
3. If the unit is not overheating, suspect a specific thermal-sensitive component (see the companion fault-tree for the full branch walk).
4. Test in the actual hot failure window, not on a cold arrival check.

References

• NFPA 70B, recommended practice for electrical equipment maintenance
• Manufacturer nameplate values for capacitor microfarads and rated current
• See related: Works Cold, Fails When Hot: A Thermal Fault Tree; Cold-Start Faults vs Warm-Start Faults: The Real Difference; Fails Cold, Runs Fine Once Warm (Decision Tree)