E.g. users at the time anyway, had massive problems with dropdown lists (a substantial number of errors were down to users who had clearly started typing, and got what the dropdown moved to as a result of them typing out the correct value). We ended up avoiding them (expiry dates in our billing form, for example).
(This was more than ten years ago, though, so our specific finding might very well be invalid now, with people more used to websites)
I'm not sure how this type of validation would play out. On one hand, if done well, it gives people immediate feedback. On the other hand it gives you no hints about what causes common mistakes like that. I like the idea of logging the values on error...
I can envision that could be easily to push this to Sentry https://getsentry.com/welcome/ using the Sentry Javascript Client
Other comments talked about client-side validation being mainly for UX and user convenience. But I'm wondering if it has a noticeable effect on whether or not people actually submit the form all the way to completion (i.e. no errors).
I made a tool that I have server-side form validation. I found that saving the invalid input so that I could examine it later was extremely valuable in understanding what people were entering, so that I could make the tool smarter. A good example is like with Parsley's demo where it requires http:// for the website URL. After seeing that people tried to enter URLs without http://, I knew I could change my code to automatically infer it. http:// is obvious, but a lot of times, people will surprise you. I don't have stats, but presumably this kind of change increases conversion. I realize this anecdote is a different effect than what I'm asking about client-side validation.