There is a not-so-smart workaround for this problem. However I want to first elaborate on this problem a little further so it gathers more attention on this site.

In multi-pulse nmr experiments, the easiest way to introduce an effective field is to simply go off-resonance. This of course requires the pulse sequence to have a nonzero scaling factor. WHH4 and MREV8 are such pulse sequences, with respective scaling factors of 1/sqrt(3) and sqrt(2)/3.

How does one introduce an effective field without going off-resonance? If the effective field is to be generated along Z, one way to achieve this is to rotate the pulse sequence, i.e., increment the phase of every pulse in the sequence. If we rotate the first cycle by an angle phi, the second cycle by an angle 2phi, and so on. We can generate an effective field of phi/(2 cycle time) along Z.

If one works out the math, it can be shown that in order for this method to work, one needs to change the receiver phase after every cycle. This is where the problem lies for stroboscopic acquisition.

For stroboscopic acquisition, at least in XWINNMR/TOPSPIN, I have not found a way to change the receiver phase after each detection. The receiver phase seems to be fixed with the adc command, and can only be changed after each scan, but not during acquisition in a single scan.

workaround: we can manually do the phase correction on the data to account for the receiver phase mismatch. This is of course, less preferred than directly changing the receiver phase during acquisition.