I guess I have found the culprit in the bad placement of formulas without an equationnumber.
The minimal example I used is that found in strc-mat.mkiv, added a [-] in the middle one.
\setuppapersize[A6][A6]
\setupformulae[spacebefore=0pt,spaceafter=0pt]
% <HERE REDEFINED \dododoformulanumber>
\starttext
\placenamedformula[one]{first} \startformula a = 1 \stopformula \endgraf
\placeformula[-] \startformula a = 2 \stopformula \endgraf
\placenamedformula {second} \startformula a = 3 \stopformula \endgraf
\stoptext
It appears that in the case of an absent equationnumber extra stretching takes place at the position of the (now absent) equationnumber. Placing phantom ()'s remedies this. The effect is best seen by enabling the \rlap{\tracedformulamode}%-line.
May I ask Hans to take a look and change it in the next beta if this indeed a correct solution? For my suggestion for the change see below.
Thanks in advance.
Hans van der Meer