Fine delay is set to 1MHz output 100ps pulse:

fmc-fdelay-status -d 0x17
Input    : enabled
  timestamping:              enabled
  50 Ohm termination:        disabled
Channel 1: pulse generator mode (triggered) 
  start at:         1643036937:000,000,000,000 ps
  pulse width:               0:000,000,100,000 ps
  repeat:                    infinite
  period:                    0:000,001,000,000 ps
Channel 2: disabled
Channel 3: disabled
Channel 4: disabled

Data acquisition

$ ./fmc-tdc-tstamp -D 0x19 -c1 -s 1000000 > /tmp/data2.txt
$ wc -l /tmp/data2.txt 
3000000 /tmp/data2.txt                                                    <----- 3x lines than samples, which is ok
$ time ./fmc-tdc-tstamp -D 0x19 -c1 -s 1000000 > /tmp/data2.txt
^C
fmc-tdc-tstamp: killing application

real    0m18.874s
user    0m0.623s
sys     0m0.871s
$ wc -l /tmp/data2.txt 
1530000 /tmp/data2.txt                                                   <---------------- should be 3000000 lines

Further commands for the same TDC gives no output at all:

$ time ../3/fmc-tdc-tstamp -D 0x19 -c1 -s 10
^C
fmc-tdc-tstamp: killing application

real    1m2.139s
user    0m0.004s
sys     0m0.095s
(tdc) [acc-py] root@cfc-774-cdv33:/nfs/cs-ccr-nfshome/user/awujek/tdc/pytest>wc -l /tmp/data2.txt 
0 /tmp/data2.txt

The described behavior above can be due to the buffer overflow. However, in such case I would expect that samples are lost, but the acquisition is not stopped.