The Model 1 was a self contained computer that plugged in to, essentially, a TV set. And that was all well and good, if feature light (16K RAM, Cassette port, limited text and graphics). But it was plug and play.
The original version, w/4K and Level 1 BASIC was notorious, for example, for key bounce. You could buy a simple mod to fix it, and it was remedied in the Level 2 BASIC.
But then came the Expansion Interface. And the EI plugged in via a ribbon cable to the main unit, offered up a printer port and diskette interface, as well as RAM expansion (it may have offered a serial port, but the one we used never used it, so I just don't recall).
Ribbon cables from the main device, ribbon cables out to the Disk drives.
Ribbon cables are very good antennas. Plugging antennas in to computer boards just doesn't work well. So, these were a glorious source for adding noise to the system from, just, well, anything near by. There was one example of a computerized toy tank that when it traveled underneath a TRS-80, the tanks CPU noise would crash the TRS-80.
Undoubtedly the original TRS-80 was part of the reason the FCC got in to the business of regulating computers. The FCC regulations is a primary reason the Model III replaced the Model I.
All the spurious noise just adds uncertainty to the system. Our particular system had no problems with it, and was pretty much rock solid reliable, though we did have some unremembered issues with the disk drives every now and then.
But one thing, from a design point of view, that was particularly striking of the TRS-80 was something my father stumbled upon as he was looking to interface hardware to the machine for custom projects.
The Z80 has two pins. A Bus Request pin, and a Bus Acknowledgment pin. Outside electronics send a signal the REQ pin, and the CPU will essentially "finish what it is doing", go in to a sleep state of sorts, and then trigger the ACK pin to let the requester know it is done.
Another thing the TRS-80 had was a series of tri-state buffers, notably, interfacing RAM to the CPU. Without getting too technical, the tri-state buffer lets one set of electronics, roughly, disconnect from other electronics in the same circuit. So, these were a mechanism to isolate the CPU from the RAM.
But on the TRS-80, the tri-state buffers were tied to the Bus REQUEST pin, not the Bus ACK pin. That means, when someone from the outside wants to ask the CPU for the bus, the first thing it does is disconnect the CPU ... from the bus! So, it's very possible that the CPU can't "finish what it's doing". For example it could be reading a multi-byte op code (like, a JP instruction), and after reading the opcode, the RAM just vanishes before it gets the actual address. The bus request lets the Z80 finish things like that.
But not on the Model 1. That RAM gets yanked out from underneath it before it can say "Wh...".
Very strange design decision. Naively, you would think they would tie the tri-state toggle to the bus acknowledge pin.
This was the wild west, folks. Don't be amazed at the problems you see, rather be amazed this stuff worked at all.
All that said, we were happy with the TRS-80. When I was getting my own first machine I was debating between a Model III and an Atari 800. I ended up with the Atari. Good friend of mine loved his Model III, however.