1: \begin{abstract}

2: Let $G$ be a linear algebraic group over an infinite field

3: $k$.

4: Loosely speaking, a $G$-torsor over $k$-variety is said to be versal

5: if it specializes to every $G$-torsor over any $k$-field. The existence

6: of versal torsors is well-known.

7: We show that there exist $G$-torsors  that admit even stronger

8: versality properties. For example, for every $d\in\N$,

9: there exists a $G$-torsor  over a smooth quasi-projective $k$-scheme

10: that   specializes to every

11: torsor over a quasi-projective   $k$-scheme after removing some codimension-$d$

12: closed subset from the latter. Moreover,

13: such specializations are abundant in a well-defined

14: sense. Similar results hold if we replace $k$ with an arbitrary base-scheme.

15: In the course of the proof we show that every globally

16: generated rank-$n$ vector bundle over a $d$-dimensional $k$-scheme of finite

17: type

18: can be generated by $n+d$ global sections.

19: 

20: When $G$ can be embedded in a group scheme of unipotent upper-triangular

21: matrices, we further show that there exist $G$-torsors specializing

22: to every $G$-torsor over any affine $k$-scheme.

23: We show that the converse holds when $\Char k=0$.

24: 

25: We apply our highly versal torsors  to show that, for fixed $m,n\in\N$

26: and a fixed semilocal $\Z[\frac{1}{n},e^{2\pi i/n}]$-ring $R$ with infinite residue fields, 

27: the symbol length of  Azumaya algebras over $R$ having degree $m$ and period $n$

28: is uniformly bounded. 

29: Under mild assumptions on $R$, e.g., 

30: if $R$ is local, the bound depends only on $m$ and $n$, and not on $R$.

31: \end{abstract}

32: