Yes, we should revisit comparisons. I want case-independent comparisons, which we have except in names of things (variables, blocks, etc.). You want a sensible treatment of strings of digits. If it were up to me, <is __ a [number] ?> and <is __ a [text string] ?> would both be true for numbers in either form. That's a hard sell because until now (almost) everything has a unique type, but I think it's the right thing. I don't think that should require being turned off and on; it's just the right thing. The fact that we don't quite handle numbers correctly is a bug, not a feature.
The trouble is, Jens complains every time we do anything different from JavaScript in dealing with types, because it requires adding tests (if/then/else) to the implementation, which slows it down. But since hyperblocks, Snap! is so amazingly fast that maybe I can convince him, supposing we ever have the time to do a serious big-picture design review.
We haven't yet had the discussion I think we should have, and so I don't know for sure what Jens will say when we do. But based on past similar conversations I don't think it'll be an easy question to settle.