This is not about what is "better" or not. It is a question whether the source of signal has low enough impedance or not.
In many cases the oscilloscope probing is meant to be "non-invasive", and not alter the signals. So if the signal have relatively high impedance, even 50-65 Ohms (as many modern signals on PCBs do), the scope probe must have an input impedance much higher than that. In some cases even 500 Ohms probe tip can be considered "high impedance probe", and there are "passive" probes that are 10:1, and look like a simple coaxial cable, see for example the P6150 Tektronix "probe". This probe is essentially a high-quality 50-Ohm coaxial cable with a tip that has a 450 Ohm series resistor, which forms 10:1 divider into 50-Ohm cable:
The resulting input impedance is just 500 Ohms, but for a 50-Ohms signal source it changes the signal just by 10%, which is frequently good enough. The major advantage of this kind of probes is that they are limited only by quality of cable (and bare input bandwidth of the scope), and P6150 works up to 9 GHz.
If you have signals with 2-3 kOhms or more at source, you must use more sophisticated active probes (if the range is 500 MHz ++), or passive probes, with mega-ohms impedance range if your signal frequencies are DC or low MHz.
However in some cases of devices the signals are straight 50-Ohm outputs, and they are supposed to be loaded with 50-Ohm loads for normal operations. In this case a direct cable connection to 50-Ohms coaxial scope inputs can and will be used. An example would be testing of signal quality (amplitudes, jitter, eye diagrams) on USB 3.0 channels (from USB.org):