A cell, yes. But not a panel that had a bad cell (for most panels with bypass diodes.)
Consider a 150W "24V" panel, which might be about 36Voc, 30Vmp, and has three diode-bypassed sections.
If Isc = 5.5A and Imp 5.0A, each section is 12Voc, 5.5A Isc.
At maximum power point it puts out 30V, 5.0A. Voc = 36V, Isc = 5.5A
In the event one section had a cell broken such that it only delivered 2.5A:
At 30V it delivers 2.5A, 75W
Open circuit, Voc = 36V
Short circuit, the two good sections push current past bypass diode of bad section, Isc = 5.5A
This panel would deliver maximum power at 20V, 5.0A, 100W
Only some SCC would move past the 75W peak to find higher wattage at lower voltage.
Same issue with partial shading, shading of parallel strings, shading of half-cut panels.
So only measuring under load is that detected. If you look through tables of measurements in my thread, I highlighted bad ones with colors. Two panels delivered less than half the power that good ones did. Their Isc was reduced, but not by that much. Possibly, Isc was only reduced due to bad (but not as bad) cells in other sections. If just one cell bad, possibly Isc would not have given any indication.
I've thought the degradation could have been due to bias voltage away from ground. Unfortunately I didn't keep track of series connection order while disconnecting for testing (wiring was interleaved to minimize use of extension wires.) The panels spent most of their lives positive biased in a negative-ground array, but the last year before these tests in an ungrounded array with transformerless inverter. That meant half were biased negative, and may have suffered PID. This could have affected all cells to some degree, although those closer to the frames maybe more so.
It is not unlikely that some batches of panels are on the used market because they suffered degradation. Others we read are because part of a solar farm was damaged, and the surviving panels were offered for sale. Caveat Emptor, which is what testing is about.