I would like to bring up something for consideration that's really come from the discussion in the comments on Benjamin Dickman's answer.
There is a problem with the dichotomy presented in the question:
my choice boils down to this: should I drop questions from my exam, even though this means fewer topics on the exam, in order that more students will not experience time-pressure during the exam?
In Norway we seem to have done our best to do away with time pressure. Our exams are 4 hours long and are not (in the lower level courses) designed to take anywhere near that length of time. Nevertheless, it is rare that students leave before the time is up and it is still common to see "I ran out of time" written on exam scripts. So simply making the exam shorter does not remove the time pressure. It also has some perhaps undesirable side effects. Since it is now feasible for a reasonable student to complete all the tasks, a good student is expected to do so and to do so well. This makes the exam even more sensitive to "computation error" than before since a good student cannot compensate for making a few such errors by completing more questions. Sometimes this is appropriate, but not always. So shortening the exam not only does not completely alleviate time pressure but also increases the accuracy pressure.
An alternative is that which is (or was - my experience is now approaching 20 years ago) practised by Oxford University. The exams were designed so that it was impossible to complete all the questions. Even just writing out an answer to a question without having to think took about half an hour, so in a 3hr exam the most even the best students could do was about six questions. There were nine on each exam.
To emphasise the point: in my year, the best student's best papers consisted of 6 good questions (given that the best of these consisted of Representation Theory and $C^*$-algebras, it's a bit confusing why he's gone in to Differential Topology, but there's no accounting for taste).
This had the effect of replacing time pressure by choice pressure. You now had a choice as to which questions to answer, and thus could, to a limited extent, compensate for lack of ability in one area by excelling in another (of course, the questions were designed so that you couldn't get away with being completely ignorant of the core). This was exaggerated by the fact that the final grade was cumulative across all exams, not just one; but I recognise that that is not possible in the modular system practised by most universities.
The emphasis on choosing questions to do well on was further strengthened by the exotic mark scheme that Oxford used (sadly, I hear no longer) which put far greater emphasis on completed questions than on partial questions. So one complete question was worth four half completed questions.
Overall, the purpose of this type of exam is to give the good students sufficient space to show that they are good, whilst giving the excellent students room to show off, and giving the less able students every opportunity to show that they have learnt something.
In a system where you don't have complete freedom to revise the overall grading system, this could still be implemented in a course simply by revising the grading scheme for that exam. The idea is to make sure that it is not out of 100 and to not scale the points. So an A is not 90% but is some number of points (actually, I wouldn't go for a point scheme at all, but that's a different issue), and a B is not 80% but is some slightly smaller number of points.
In the Oxford system, in my year, then one of the criteria for a first class degree was (roughly) to get at least 20 good questions and 20 fairly good questions. That's spread over 13 papers, so that's just under 2 good and 2 okay questions per paper. That's from a total of 9 questions per paper, so there's plenty of room there for a candidate to choose their questions wisely, and plenty of time for a good candidate to achieve that level.