The xlimit of the runtime profiles should probably depend on the data. Runtime is to a great extend scale invariant and would in practice be multiplied with the time per evaluation which has an extremely wide variation depending on the function, like easily 7 orders of magnitude. Hence, the x-scale and -limit are to this extend "arbitrary" and function dependent anyways.
The horizontal distance between graphs is the relevant measure and would then change on a per-figure basis. However, the grid will remain the proper guide to extract quantitative differences and this is easier to read with a tighter x-limit.
The xlimit of the runtime profiles should probably depend on the data. Runtime is to a great extend scale invariant and would in practice be multiplied with the time per evaluation which has an extremely wide variation depending on the function, like easily 7 orders of magnitude. Hence, the x-scale and -limit are to this extend "arbitrary" and function dependent anyways.
The horizontal distance between graphs is the relevant measure and would then change on a per-figure basis. However, the grid will remain the proper guide to extract quantitative differences and this is easier to read with a tighter x-limit.