[restalis]

"your theory has three burdens to meet:"

...

"it must make new predictions that can be experimentally or observationally tested, and confirmed or refuted, that are unique to this new theory"

("MOND fails spectacularly, either offering no predictions...")

I fail to understand the necessity for new predictions. This simply may or may not come out of new theories as a paradigm shift consequence and it is therefore just optional, not a must. In fact, I fail to fully understand even the first requirement (about reproducing all the successes of previous leading theory) - the current leading theory's explanation may happen to explain things somehow, but that may not be necessarily correct. I wouldn't dismiss a competing theory even with holes in it if there is a slightest chance that further work on it or on something closely related to it may come to cover those shortcomings in time.

[drpre]

Disclaimer: far from a physicist.

I think the point is that, for a new theory to become more favored than the current leading theory, the new theory must be consistent with at least as much scientific evidence as the leading theory. From there, the value of predictions is that they give a path for strengthening the theory further.

I think you're right in saying that these are not hard-and-fast rules.