I thought we all agreed long ago that sockets are evil... Like you said, momentary disconnects cause instant death, which is why we've all sworn off sockets long ago... If that diode wiggles a little in the socket, poof, instant LED.
Odd advice, considering how long you've been here, or, do you know something about switzerland and their sockets that we don't?
It's interesting how sockets have gotten such a bad rap, after all, processors and chips use them all the time without problems. In fact, you can buy laser diode
sockets from companies such as Thor Labs and they appear to perform well enough. People "agreed" but did they do any actual testing?
What I suspect is that people simply used the wrong headers for the job. Remember that headers/sockets are
designed to prevent momentary disconnects. You don't want, for example, the chip you've connected to suddenly losing its connection during a computation. This is why sockets are relatively expensive: they're designed to
perform like a soldered connection, but enable you to replace the device without desoldering.
Where a header or socket may
fail you is if it's the wrong size for the pins; so it doesn't properly grip the device. It can also fail if there is significant mechanical stress on the connection (something pulling on it). A "wiggle" shouldn't be significant, and would cause a disconnect only if there was a size mismatch between the pin and device.
"Standard" headers are designed to connect to other headers. The male headers are long and square, while the female standard headers are made using a "fork" of metal inside a plastic casing that grips the male head. These kinds of headers are completely inappropriate for connecting to devices directly because of their large size. That's why you put an IC in a socket and then put the socket into the headers. I've tried using them for replaceable LED sockets, and they're not good for that.
The (Swiss) machine-pin headers I'm suggesting for diodes are meant for connecting to devices such as chips, instead of
just other headers. They are smaller than standard headers, and not only that, round and often plated with brass. Female machine-pin headers grip the male variety with a completely metal hole, ensuring a good connection and strong grip. The female header holes are exactly the right size for the pins on a laser diode, and I suspect that these are the exact same socket connectors used in Thor Lab's laser diode sockets.
My claims aren't based solely on conjecture. I wasn't sure if they'd work either, so I bought some to test. How do they perform? They have a very solid grip that would take effort to disconnect, and are also spaced the same as the pins on a diode, providing the set with structural support. The most important question is, however,
would they cause momentary disconnects? Not unless you're actively pulling the diode off the socket, which would probably damage a solder joint as well. After the diode is installed there shouldn't be much mechanical stress on the system anyway, and if you're afraid, you can always hot-glue the connection down, which is a good idea anyway.
Why use it instead of soldering? Don't, if you're comfortable soldering directly to the diode after it has been mounted. Sockets are far more expensive than solder + wire, and the soldered wire is a
better permanent solution. However, the sockets
can be useful for quickly connecting to mounted diodes after the rig has been completed and mounted; for example if you've glued in your driver and don't want to mess around with soldering inside the box. They're also great for replaceable parts, and testing many diodes at a time, as Laserbee has used them for.
Do not, however, dismiss them as a solution for connecting to diodes. They
do connect very solidly to laser diodes provided you're not putting them in extraordinary circumstances. Inside a pen, or box, or other static arrangement I would definitely not fear that the socket would pop off the diode any more than a soldered connection.