While simulating the activity diagram in the figure (which is a toy example derived from a more complex one), I would like "action 2" to hold "Data A" after the refinement in the decision node happens, until a new value is produced by the action upstream. However that does not happen and the simulation just halts, with "action 2" waiting for "action 1" to produce a new "Data A".

I found 2 possible solutions:

1. To make the multiplicity of "Data A" [0..1] in the input pin of "action 2". This solution does not fit my needs because I would like to express that the input pin typed with "Data A" be granted an input at least once.
2. To put a Data Store between "action 1" and "action 2" to send the value of "Data A" at all iterations once received. This solution does not fit my needs because the Data Store accumulates values in case - in more complex examples - it is fed more than once (the <> stereotype from the standard does not seem to work here). Moreover this solution would be sub-optimal because I would like to represent holding the input fixed as part of the behavior of "action 2" rather than keeping this information external.

What is the recommended way to represent "Data A" being memorized by "action 2" for subsequent iterations (i.e. after refinement happens in the decision node)? I also tried to play with behaviors (including opaque behaviors) but that doesn't seem to take me anywhere.

I am using CSM 2021x

Thanks