I just read interesting advises about how to design good PCBs. But they apply mostly to a professional consumer electronics. PCB is not so critical for a DIY person. Maybe if your board uses Gigahertses, you should be awake when you design it.
Examples
A friend of mine left out all the power supply capacitors of his STM32L0 board. He added only those what were absolutely needed.
I once forgot to connect one power supply +5V pin of Atmega2560 CPU. The board worked fine, until there were assembly problems elsewhere on the board and I found out this too.
I once made a low power 144MHz crystal controlled FM transmitter on a perforated board (Veroboard). Power and ground were 1-2 inches apart. The thing even worked.
But still I usually route power and ground first. The rest is less important.
There is a big difference between “working” and “working properly”.
Your “friend’s” STM32 circuit may reset “unexplainably” once a month because some noise is coupled in, or you neighbor can have very bad reception on his HAM stuff because your circuit generates loads of EMI.
There is simply no excuse for making a mess of things when designing a PCB you want to use for longer times (and want it to be reliable). I also cut some corners when working with temporary circuits on breadboards (but I don’t skip on decoupling caps, especially with bad quality power supply routing on breadboards these are more important then ever).
Around 1989 we made an EPROM emulator. As those time microcontrollers were really expensive (at least for us here - for one month salary I could buy two 8748) it was done as an adapter to our EPROM programmer (so no next microcontroller needed). It contain RAM + buffers at both sides.
It worked. But from time to time (once per minute I think) it happened that byte was read corrupted from it. The only tool we had (in our firm) those time was oscilloscope I have done myself being student. It was one channel, 5MHz and of course analog (no way to catch and hold runs from problematic moment). I could’t find the source of problem. We borrowed the better oscilloscope for one day. I noticed spikes at VCC. Adding one capacitor solved the problem. Those time the microcontrollers were still new things for us. We were still learning how to use them.
Since that time if anything works but not fully properly the first what we are looking for is if there are enough capacitors there.
There are reasons that best practices and manufactures recommendations are generally heeded. Unless you REALLY know what you are doing, stray at your own risk.