In the present work we study the C8 flare occurred on September 26 , 2000 at 19:49 UT and observed by the SOHO/SUMER spectrometer from the beginning of the impulsive phase to well beyond the disappearance in the X-rays .
The emission first decayed progressively through equilibrium states until the plasma reached 2-3 MK .
Then , a series of cooler lines , i.e .
Ca x , Ca vii , Ne vi , O iv and Si iii ( formed in the temperature range \log T = 4.3 - 6.3 under equilibrium conditions ) , are emitted at the same time and all evolve in a similar way .
Here we show that the simultaneous emission of lines with such a different formation temperature is due to thermal instability occurring in the flaring plasma as soon as it has cooled below \sim 2 MK .
We can qualitatively reproduce the relative start time of the light curves of each line in the correct order with a simple ( and standard ) model of a single flaring loop .
The agreement with the observed light curves is greatly improved , and a slower evolution of the line emission is predicted , if we assume that the model loop consists of an ensemble of subloops or strands heated at slightly different times .
Our analysis can be useful for flare observations with SDO/EVE .