My understanding of the signaling model is that it depends crucially on the relative
cost of signaling to people with and without the desired trait. You
want the cost to be high for someone without the trait and low for
someone with the trait.

With that in mind, I do not see how lowering the cost of signaling
for people with the trait does anything other than cause people with the
trait to choose the low-cost signal. The problem with a low-cost
substitute for a diamond is that it lowers the cost of signaling for
people without the desired trait (which is a willingness to buy an expensive gift).

The problem is that a low-cost substitute lowers the cost of signaling for everyone.  So if the cost per signal falls by 50%, you have to do twice as much signaling to separate yourself from the pack.

Simple example: Suppose that (a) good students are $20,000 more productive than bad students; (b) good students endure $5000 of suffering  per year of school; (c) bad students endure $10,000 of suffering per year of school.  Then in equilibrium, good students need at least two extra years of schooling to distinguish themselves from bad students.  Good students will be happy to do so, because it nets them $20,000-2*$5,000=$10,000.  Bad students won’t bother, because imitating good students nets them $20,000-2*$10,000=$0.

Now what happens if the cost of education falls by 50% for both groups?  A two-year education gap is no longer stable!  Bad students will suddenly find two years of education profitable: $20,000-2*$5000=$10,000.  Now the good students need four years of schooling to distinguish themselves.  As a result, the total value of resources devoted to signaling remains unchanged.

If I come up with a low-cost way to earn a badge that signals
intelligence, conscientiousness, and conformity, and that badge can only
be earned by people with those traits, then my badge should find a

If you devise a low-cost signal that only high-ability people can earn, you’re right.  But that’s tautological.  In the real world, low-ability people can always try to imitate high-ability people.  If the signal everyone used to send gets 50% cheaper for everyone, the quantity of signaling has to double to preserve separation. 

One challenge is that when few people use the badge, it seems
to signal non-conformity. Thus, the early adopters of my cheaper badge
do not do as well as they should. But over time, there are two
possibilities. One is that the conformity hurdle cannot be overcome, so
that the incumbent signaling mechanism remains dominant forever. The
other possibility is that eventually a tipping point is reached, and
enough people use the new badge so that it no longer signals
nonconformity. At that point, the market position of the old badge
rapidly deteriorates.

I think that we will arrive at the second equilibrium at some point. However, predicting when it will occur is difficult.

What’s your best guess, Arnold?  Now you barely sound more sanguine than I do.