I think a good way to answer this question is to present what is observed experimentally. This involves a careful statement of what we mean by "speed". So I will explain this by an extended example.
Suppose first of all that we collect or manufacture 200 steel rods, all the same length. Let them be 1 metre long for convenience. We can stack them on top of one another to be sure they are all the same length relative to one another when they are not moving relative to one another.
Next let's manufacture 4 clocks, all of the same design, and make it a robust design such as an electronic circuit based on a quartz crystal or something like that. We make sure these clocks agree with one another in the length of a second and a millisecond and a microsecond etc.
Now take 100 of the rods, and lay them end to end down the length of a corridor. We then say the corridor is 100 metres long. Or, more carefully, we say the "proper length" of the corridor is 100 metres.
Also take two of the clocks and put one at each end of the corridor.
Next imagine a large space ship that can travel very fast. Take the remaining 100 rods and attach them to the space ship, laid end to end. Also furnish the space ship with two clocks. Finally, gently accelerate this space ship until it is travelling very fast relative to the corridor. Let's say that the speed of the space ship relative to the corridor is $v = c/2 \simeq 1.5 \times 10^8$ m/s.
Finally we do a simple experiment to measure the speed of light. We suppose a light pulse propagates down the corridor from one end to the other. In this experiment there are two events to think about: event $A$ as the light pulse leaves the first end, and event $B$ as it arrives at the other end. As far as observers at rest relative to the corridor are concerned, these events are separated by a distance of 100 metres. To measure the timing, we can use the clocks at each end, but we need to first synchronize them. This is done beforehand, by sending light pulses in both directions from the centre of the corridor and setting both clocks to zero when those pulses arrive.
In this experiment it is found that if the time at event A is zero then the time registered by the clock fixed to the corridor at event B is $0.33356$ micro-seconds. So we deduce that the speed of this light pulse relative to the corridor is
$$
\frac{100\;{\rm m}}{0.33356 \times 10^{-6}\;{\rm s}} \simeq 3 \times 10^8\;{\rm m/s}.
$$
Ok, now let's consider that same light pulse, but now observed from the space ship. To be precise, we imagine the space ship is zooming down the corridor at the same moment when the light pulse sets out, and we can imagine that the light pulse leaves burn marks on the set of rods fixed to the ship. Let's suppose it leaves one burn mark as it sets out, and another when it reaches the end of the corridor. We can arrange that the clocks in the space-ship are located at these two burn marks, and we can arrange for them to be synchronized for observers in the space-ship.
Now it is found in this experiment that the two burn-marks on the spaceship are separated by $57.735$ of the steel rods on the spaceship. And it is found that the clocks on the spaceship register a time difference of $0.1926$ microseconds between the two events. (I am not offering a proof, just a statement of what is in fact found in experiments like this one). So in view of these observations we say the speed of the pulse of light relative to the spaceship is
$$
\frac{57.735\;{\rm m}}{0.1926 \times 10^{-6}\;{\rm s}} \simeq 3 \times 10^8\;{\rm m/s}.
$$
In this example a single pulse of light was observed to travel between a single pair of events. The speed came out as the same
$(3 \times 10^8$ m/s) when measured by two sets of observers who
were themselves travelling at $c/2$ relative to one another.
