A high-level neutron coincidence counter equipped with 18 3He tubes and a JSR-11 shift register unit with a detection limit of 0.20 n/s for a 2-h run is used to study the neutron signals in D2 gas experiments. Different material pretreatments are selected to review the changes in frequency and size of the neutron burst production. Experimental sequence is deliberately designed to distinguish the neutron burst from fake signals, e.g., electronic noise pickup, cosmic rays, and other sources of environmental background. Ten batches of dry fusion samples are tested, among them, seven batches with neutron burst signals that occur roughly from −100°C to near room temperature. In the first four runs of a typical sample batch, seven neutron bursts are observed with neutron numbers from 15 to 482, which are 3 and 75 times, respectively, higher than the uncertainty of the background. The samples seem to be inactive after four or five temperature cycles, and the inactive samples could be reactivated by degassing and recharging of deuterium. The same anomalous phenomena were observed in theMentou Valley Underground Laboratory situated 580 m below ground.