The switches on, say, an IBM 1401 and 1620 were negligible. The lights could tell a lot about the state of the system, however.
The CDC 6000-7000-STAR, etc. had no switches or lights. There was a "deadstart panel" with a matrix of toggle switches whose contents were initially used to start the system. The operator's console was used for interaction. In fact, if one's hearing is blocked out, it's pretty much impossible to tell if a 6600 is powered on and "hung" or just powered off. The display had to be refreshed every few milliseconds or it went blank. The STAR and 7000 used "MCUs"--basically a minicomputer with its own memory and display that allowed one to manipulate all sorts of interesting things. The STAR MCU had its own drum, which was initially loaded from paper tape. Memory-and-address switches are not to be confused with "sense switches". An interesting early feature of FORTRAN was the ability to interrogate the status of (usually 4) hardware or software switches. There was a less-used feature of "sense lights". When requesting that the operator mount a particular tape, for example, the message to the operator was something like "PLS MOUNT TAPE xxxx UNIT yyy 1 UP AND GO". FORTRAN II had the "IF SENSE SWITCH..." statement; Some FORTRAN IV versions used library routines to do the same thing; e.g. SSWITCH, SLITET). --Chuck