Hi,

Thanks for the notice and I'm sorry about the frustration. I'll try to explain briefly what has happened:

First, the loss of polyhornerscheme was a small oversite, we only include the packages that are needed, so we look for the command, currently we have the following for polynorm package:

"\\longdiv",
	"\\polyadd",
	"\\polydiv",
	"\\polyfactorize",
	"\\polygcd",
	"\\polylongdiv",
	"\\polylonggcd",
	"\\polymul",
	"\\polyprint",
	"\\polysub",
	"\\polyhornerscheme"

You've see I've put back in 'polyhornerscheme', thus:

If there are others we need or think we should include just let me know.

The tabular command we have just recategorised (in the past 3 hours) to be interpreted in a strict manor. Therefore the entire equation must be grammatically perfect or it'll be rejected. The reason previous equations may work is they are cached by our servers and so long as they're access frequently they'll remain there. However similar new equation must now be written perfectly or they'll fail.

i.e.

\begin{tabular}{ l c r }
  1 & 2 & 3 \\
  4 & 5 & 6 \\
  7 & 8 & 9 \\
\end{tabular}

which gives

The problem with tabular is that its not designed for maths, so if you insert any standard maths notation into it (i.e. ) it'll fail.

You should therefore use 'array' as a straight replacement, i.e.

\begin{array}{ l c r }
  1 & 2 & 3_b \\
  4 & 5 & 6_a \\
  7 & 8 & 9^2 \\
\end{array}

which gives

To answer your final dilemma: WHY? We hate having to change our algorithm, but as we learn how our system works and also what equations break or degrade our service (albeit temporarily) we adapt accordingly. For the most part we have designed our system to be flexible about the grammar: i.e. if we know enough from what you write to produce something, then we try... Unfortunately while the LaTeX language is incredible, its also incredibly complicated and not all packages are as robust as other, so the smallest typo can leave threads hanging. Hence why were stricter over certain command than others.