Lacking points was never an issue for me. It wasn't a lack of blobs but more that they'd get stuck in awkward places. The "power to self destruct the blob, at a cost" sounds like it would be a good solution.
As for the inconsistency. It only happened on the final level, though twice in two different locations. Both times I'd send a single test block, who'd succeed, following which I'd send a bunch who all do some alternative (miss a jump or something of the sort). I think I might of just realized the issue. When I sent one blob, everything's fine. But multiple meant there may have been a subtle performance hit, adjusting certain stats in some ways, ergo leading to a different output. Furthermore, each time this happened they were all sped up, my guess is that when sped up there's a lot more room for the math to vary since it's working with bigger numbers. The final level is also very crammed with variance in level geometry and stuff so there's more chance for this sort of thing to occur.