Parallel pairs is better (that is a single series string of cell-pairs)
Consider what happens when a cell dies with high internal resistance (or has a protection circuit which switches off). In with 2 parallel strings of 4 cells, you lost half of your capacity because the other three cells in that string of 4 with the bad cell can't work either because their current can't flow through the dead/off cell, where with parallel pairs you really only lose that one cell because the other in the parallel pair can carry on working and carry the current, not quite that simple since it's current delivery capacity of that cell pair is now half, but it's not as bad as totally losing half your pack and if you're not pushing the cells to their limit in current delivery terms anyway it won't matter.
Similarly consider what happens if the cell develops a shorting fault (dendrite formation), in 2 parallel of 4 series cells you basically have the current of 4 cells able to dump through into that short, where in a 4 series of 2 parallel cells you have only the single paired cell which can feed the short. Either way the result probably won't be great but the lesser of two evils is better!
Similarly if you will monitor the cell health there is much less to monitor if you use parallel pairs, if you use 2 parallel of 4 series cells then you have 7 points to measure to get the cell health, while in 4 series of 2 parallel cells you have only 4 points to measure to get the cell-pair health.
Laptop packs invariably use parallel pairs in series.