I would guess both. Yet I also think that the specific time frame is critical.
When using a short 30sec forced swim as a contextual reminder for a previous near-drowning experience we found that hyper-mobility, and 'behavioral coping' time (attempting to 'wall climb') in particular, were significantly increased (as well as strongly related to individual degrees of amygdala activation).
We are all aware about all problems that behavioral tests carry as well as their interpretation. Original Porsolts test requires 2 swim session, 24 h apart. 90% of papers published up to now identify FST as a test to assess antidepressant response, or depressive-like behavior after exposing to some stresor or as a type of stress model. But recently, I come a cross several papers that question these paradigms with very sense comments. Taking into consideration the authors that published these reviews these information's are of great values on one side, bu on another they question a great amount of publication and results also.
Recently, I and some colleagues were concerned with the use of single parameters to analyze behavior, since the interpretation of complex abstract behaviors (such as depressive and anxious behaviors, for example) may be imprecise. As you said, increased immobility time in the forced swimming test could indicate an increase in depressive behavior and also a greater adaptation or habituation (related to cognitive aspects). In addition, such data could be extrapolated for high muscle fatigue too.
According to Gilad's comment, we think that the most feasible way to minimize imprecise interpretation is to evaluate multiple behavioral data in multiple behavioral tasks and analyze them in an integrated way. We are conducting Principal Component Analysis (PCA) to achieve this goal.
Thus, if, after performing PCA, immobility time is related to anedonic behaviors or other depressive-like behaviors, we could say that we are measuring depressive symptoms. However, if immobility time is related to memory or learning behaviors, for example, we would say that we are measuring memory-related cognitive behaviors rather than depression.