The reason why we stopped liking paragraphs was because everyone bitched if their vote got rejected. There were a small handful of people reading these essays and all they got in return was crap. So nobody liked writing the paragraphs and nobody liked judging them. I really don't see that changing if we go back to it, and that's ignoring the obvious potential for bias in the system.
My name's been thrown around in this thread a few times now so I thought I'd weigh in. Nobody liked writing paragraphs because most people felt it an inconvenient burden to be asked to articulate their thoughts on a suspect. Most of those who did and did not have their submissions accepted bitched about it on one forum or another, which was one of the few things I didn't like about judging paragraphs because it was a constant. There have been several improvements suggested or put in place to the original bold voting system that would make or have made it easier for both parties:
1) Expanding the pool of judges
As much as a chore writing paragraphs was for would-be tiering contributors, imagine how much effort and attention it took for two judges to read dozens of no-word-limit submissions, two and sometimes three times over. Manaphy had over 70 submissions PMed to Aeolus and myself, and if I hadn't reread most of them myself, Blue Tornado and Minato would have gotten away with plagiarism. The evaluations themselves took about a week since besides the fact that Aeolus and I were generally busy, we were only two people, and for the Latios Stage Aeolus was completely out of the pocket and I had to read them all myself.
As level-headed and even-keeled as I think I am, I would have appreciated additional eyes if only to somewhat quiet the cadence of impatient foot-tapping that was often directed at Aeolus and myself. With five people now, the bandwidth of judges is increased dramatically. Lags in evaluation time and, more importantly, the risk of inefficient judging owing to a bandwidth deficiency are (or should be) a thing of the past.
2) Sentences instead of paragraphs
One of the last tests that Aeolus and I administered put in a 400-word limit. I don't think anyone had a real problem with this, and going even shorter or much shorter at least puts everyone on a more equal writing field.
3) Publishing all submissions after a given stage or vote
I was open to this whenever it came up after some votes and even acquiesced on at least one occasion. This should be mandatory if whispers of bias or faulty judgment are still seeping through the grapevine.
4) SEXP for "potential for bias in the system"
As you are one of the few proponents of SEXP (thanks) you likely understand that its introduction was facilitated entirely to reduce bias, and make the ascertainment of an "ideal voter" as automatic and objective as possible. This was not met well by the majority of our community since they felt that not knowing the metrics and components of the SEXP formula precluded its fairness even though X-Act himself stated that the formula would not work if people knew how to manipulate it, but good luck getting anyone to come back around on this.
In sum, there are some strides that can be made to improve the process as it is, and some steps that were taken in stages past should not be forgotten altogether.