Dear dr. Grover,
I think that you have answered your own question by assuming the existence of different subcellular compartments expressing different SERCAs (exhibiting opposite affinities for calcium) and calcium-release channels. One piece of evidence has recently been provided by dr. Muallem's group (J. Biol. Chem, 272, 15771-15776), who showed that SERCA2b is exppresed at the luminal pole of the acinar and duct cells of submandibular salivary glands, whereas SERCA3 was found at the basal pole. Moreover, in our lab, Baba-Aissa (unpublished observation) suggested that in rat cerebellum, SERCA2b was separated with both the rough and smooth ER membranes and SERCA3 with the smooth ER.
Regarding the co-localization of SERCA2b, SERCA3 and phospholamban in a given non-muscle cell type, there is no evidence that I am aware of it. If I am not wrong, it has been shown that phospholamban is expressed in muscle cells, but not in non-muscle cells.
Regarding the putative cellular areas susceptible to reactive oxygen species, I wonder whether it is correct to assume such effect, since these species possess a high rate of diffusion. From your experiments, it seems that SERCA3 is more resistant to oxidative stress than SERCA2b. Therefore, I like the idea that SERCA3 is advantageous for cell survival. An interesting model cell would be the SERCA3-expressing macrophage, which produces a lot of toxic NO needed to kill bacteria.
Concerning the last question, one can only speculate about the physiological implications of two SERCA pumps with different calcium affinities. One can think that a pump with a high affinity (SERCA2b) is needed for the restoration of the cytosolic calcium concentration to resting or close to resting levels after cell stimulation. A pump displaying low affinity for calcium such as SERCA3 would be beneficial for those cellular events such as secretion, which require high calcium concentration. Another possibility vehiculated in the field would be that the activity of SERCA3 might be regulated in vivo by an yet unidentified modulator like phospholamban for SERCA2a and sarcolipin for SERCA1.