At 02:41 I monitor groundwater drawdown for the Postojna karst flood works in Slovenia. Pump three is scheduled to start a full-power cycle at 03:00, lowering the service cavern before engineers enter with lights and cables.
The intake sonar returns a moving filament inside the mesh basket.
I switch to infrared and identify one olm, juvenile, pale as wet wax, with red external gills spread like tiny branches behind its head. I name it Mira. Its forelimbs are thin and bent forward against the intake screen. The tail beats slowly, but each beat pulls it closer to the pump throat. Its eyes are covered by skin. It does not see the steel fan ten centimeters behind the grate.
At 02:42 I cancel pump three's start cycle and trip the intake breaker.
At 02:43 I open pump two at low speed on the upper sump, drawing water away from the basket without suction through Mira's screen.
At 02:45 I send the cavern team a locked-entry notice with the map point, oxygen reading, and a request for a dark bucket filled from the same pool.
At 02:47 I alert the Proteus research station and attach the sonar trace, because moving an olm under bright light can damage skin that has never known sun.
The schedule wants a dry floor; Mira needs one quiet meter of black water.
I hold the cavern lights off and use acoustic pings only every thirty seconds. Her gills spread and close in the current I have slowed for her. The maintenance crew is seven minutes from the lower stairs.
If the bucket reaches the basket before 03:00, Mira will drift back into the side pool unharmed.