Workers at the nuclear power plant in LaCrosse, Wis., discovered one day in 1969 that water in one of the plant drinking fountains was slightly radioactive. It turned out that the fountain had somehow been connected to a 3,000-gallon tank of radioactive waste water.

The year before, workmen had needed something to plug a pipe while they repaired a pump in the "swimming pool" shielding around used radioactive fuel at Shippingport, Pa. The handiest thing: an ordinary basketball. Wrapped in tape, it was shoved in the opening.

Unfortunately, there was too much water pressure in the pipe, and the basketball popped loose, and 14,000 gallons of water spilled into the basement.

In a certain kind of design found in six plants in 1974, switches intended to float and close a circuit in an emergency were found to sink like lead instead. Their failure could have meant that a reactor would not have been able to shut down if necessary.

These are samples from the personal "Nugget File" of Dr. Stephen H. Hanauer, an assistant director and the senir technical safety official of the Nuclear Regulatory Commission. For the past 10 years he has collected unusual and/or serious reactor safety incidents as they passed across his desk, assembling them into a foot-thick file.

The Union of Concerned Scientists, a Massachusetts-based group of scientists critical of nuclear power, obtained Hanauer's Nugget File under the Freedom of Information Act and made public yesterday a 95-page book of excerpts.

The book, said UCS leader Robert Pollard, "will help the public understand what really goes on at nuclear plants so they can better judge the vague, soothing reassurances they normally get from the industry and the regulators."

Hanauer said he had used his Nugget File for years and is still adding to it almost weekly. "I wanted to save the incidents I could learn from and refer to." He learns, he said, "ways in which things can fail that maybe I hadn't thought of before." Some requirements for electrical system components had been changed because of his Nugget collection, he said.

"None of these are scary. Some are systems that should have worked that found ways not to work... machinery will break and people will make mistakes, but there are enough safety protections and enough redundancy to make sure that nothing [major] has ever happened."

The basketball incident, he said, has occurred several times in various forms. A plywood and rubber blocking device used during welding operations was left in a jet pump inside the reactor vessel of the Dresden III plant in Morris, Ill., in 1971, causing coolant flow to be reduced.

In 1972 at the Quad Cities II plant in Cordova, Ill., it was an entire welding outfit, complete with tanks, hoses and so on, that was left in the jet pump.

"These things happen... one time in a hundred you forget to take the things out," said Hanauer.

He admitted to "exasperation" at some mistakes. It was discovered at the Maine Yankee plant in Wiscasset in 1973 that some reactor safety circuits were so badly designed that failure in one could cause failure in all four backup systems. "Some day we will all wake up," Hanauer wrote in the margin of that report.

He said yesterday he had been referring to a continuing problem with that particular design, and not to nuclear power in general.

The Nugget File book makes it plain that problems plaguing the average homeowner also occur in nuclear power plants. Parts are installed upside down, backwards, sideways or in the wrong place. Varnish and paint goes on too thick and rods or pumps stick. There are frozen pipes, clogged drains, blown fuses, loose screws and bad soldering jobs; short circuits and unplugging plugs; dials turned to "drain" that should be on "fill" or "off."

Some of the incidents had serious consequences.

A fire broke out in the plutonium handling area of the Rocky Flats, Colo., weapons plant in 1969, burning in the radiation shielding material for more than four hours because firemen were afraid to use water to put it out. Water could cause a nuclear chain reaction in weapons-grade plutonium. Damage was estimated at $45 million and plutonium contamination occurred throughout the building as firemen tracked the dust around.

Although reactors are supposedly earthquake-proof, one 1971 accident at the Nine Mile Point I reactor in Scriba, N.Y., involved switches that were "accidentally bumped." They tripped open and caused water to overflow onto hot steam lines. "Of course, we don't know how hard they were bumped," Hanauer said.

A workman at the H. B. Robinson plant in Hartsville, S.C. was trying to vacuum clean the steam generator in 1974 and opened the vacuum cleaner to find out why it wouldn't work. It turned out to contain radiactive cobalt dust, which he inhaled, receiving enough contamination to be temporarily relieved of duty.

The files showed several incidents in which plant workers ignored flashing lights or warning bells because their instrument panels had been crying "Wolf!" with malfunctions so often that the personnel had stopped paying attention. One safety system at San Onofre I in San Clemente, Calif., had been inoperative for 20 months before it was spotted in 1968. Key valves at the Quad Cities II in Cordova, Ill., may have been wrongly open 4 1/2 months before they were discovered in 1972.

"The control panels still need improvement," Hanauer said, "but they're a lot better than they used to be."