Not Fungus, but my $0.02.
Under what conditions do you feel having a separate on/off switch makes a handheld multimeter less safe than one where the function is built into the rotary selector?
I do not think it is a big deal. The meter should be safe regardless of the switch and leads configurations. So powering on a meter with the switch set to current should not be a safety issue. That this is almost certainly not the case with this particular meter is a different issue.
Of course it can blow an (expensive) fuse or trip a breaker, shutting down a mains circuit. But those are not safety issues (unless there was a pre-existing one like a locked door that cannot be opened without electricity). I do not think it is an amazing design in combination with the shared current jacks. But I remember bench meters that used ganged switches for function switch (so it would be persistent across power off). They had a separate power button. And some had a shared V/mA socket. Never had any problems with that.
Do you feel the separate mechanical on/off switch is less safe than an electronic one?
Since power on/off on a DMM will not usually affect safety and function selection or voltage at/current through the test leads, I do not think this matters at all. Meters that use soft switching for functions (e.g. any modern bench meter) will usually default to a high impedance (e.g. DCV) state on power on.
I figured Fungus brought it up and could explain his reasoning. I appreciate the input if others want to chime in on this as well.
I've never seen a meter where the function or on/off switch has anything to do with the high current inputs. If you have the leads connected and you attach across 2 phases of your 480 bus, at best the HRC fuse will go. Worst, well for the meter Fungus refers to, there is no fuse. Still, is has nothing to do with how the on/off is implemented in the HF meter or any other that I am aware of.
Based on his initial post
The convenient on/off switch means the selector can literally be in any position when junior connects the probes.
it seems this is concern but I can't come up with a case were it would increase your risk.
Of the meters that I have remaining, the ones with a separate on/off switch are:
CEN-TECH 90899, mechanical (basically the meter Fungus refers to)
UNI-T UT181A, electronic
Gossen Metrawatt Ultra, electronic
Fluke 97 scopemeter, electronic
The Gossen is an interesting case. Forgetting that the relays can change their states from a simple magnetic hanger causing the meter to display very low voltages when high voltages present, the reason for the relays is to be able to remotely control the meter. If you were using the meter with the RF link enabled, the meters function can be changed by the remote. I really like the CEM with it's RF link and have used it. However, in every case I set the meter up then forget it. I see no value in being able to change the function from an RF link and can see this adding risk. Removing this feature would have solved the state change problem as well. I never did hear anymore out of Gossen.