Goldrick, Runnqvist and Costa (2014), summarized by Psychological Sciences.org, conducted an interesting study where they had bilingual subjects switching back and forth between English and Spanish (their dominant language) nouns. Spanish consistently influenced their pronunciation of English consonants but English did not affect Spanish consonants. Spanish influence was not readily apparent when English terms were articulated consecutively.
The point of the study is that the additional processing load, itself, of switching--not just the differences in articulation of the L1 and L2 consonants--was contributing to the emergence of the more salient Spanish influence.
In an EFL class, where more of the speaking is in the L1 the "switching-processing" effect may be quite "pronounced." Even in an ESL class, where students, themselves, may be using the L1 privately, outside of the flow of the class, the effect on L2 pronunciation could, likewise, be significant. In the structuralist-audio-lingual period, exclusion of L1 in teaching was, indeed, a given.
Now if all the switching effect does is allow in a bit more "accent," then that may not be all that problematic, but that is not what the study seems to be implying: switching causes a generalized processing overload that probably affects much more than just pronunciation. It may be time we reexamine that effect, at least in terms of pronunciation teaching in integrated classroom instruction. The study deserves replication/extension to current methodology--and a closer look at L1 and L2 switching in your class as well?
Goldrick, M., Runnqvist, E., & Costa, A. (2014). Language Switching Makes Pronunciation Less Nativelike. Psychological Science, 25 (4), 1031-1036. DOI: 10.1177/095679761352001