This is true, but can be turned to advantage
Pronunciation can be in the background of every activity, and foregrounded as it emerges in the course of any activity. To do this you need an immediate and ready to hand approach to dealing with whatever pronunciation issues come up, rather than having to make a note of a problem and then come back to it later when you have prepared something. The problem with doing it later is that the teachable moment has passed, the context is no longer live, the connection is lost, and to revive it requires new energy.
Work on it now or later?
There is more value (I think, and you may disagree) in doing it on the spot when it occurs, when the need is apparent, as there is more energy in it and it is part of an ongoing process, the need is directly at hand, and by doing it immediately you constantly confirm that if they can’t say it intelligibly there is little point in knowing it. In other words you gradually establish in your class culture the idea that an acceptable pron is an integral part of knowing a bit of language, not an extra bit of knowing. It is not a bonus, it is part of the core. If you can’t say it intelligibly then you don’t know it. Of course you can’t deal with everything, or even very much, on the spot. But we do need a focus of the moment and a way of not letting it all slip through our fingers
A vocab lesson
Saying a word and getting its stressed syllable, and more importantly its unstressed syllables, is key to saying and knowing the word. You don’t know a word if you can’t say it comprehensibly, and probably you can discern a word less reliably with your ear if you can’t actually say it. Ear and mouth educate each other. When practising vocab the teacher should keep gently demanding the best pron that the St can do at the moment. So that during the memory processing and the formation of the synapses the St does not default to their habitual L1 phonetic set. Pron is part of the identity of a word, in fact pron is the physical reality of the word and contributes vitally to hooking it into the memory.
Prime the memory for pronunciation
It doesn’t matter that the pron is not at first perfect, what matters is that at the time of learning something about pron is included in the memory synapse, the memory is somehow primed for pron, so if the memory synapse could speak it might say “…Ah yes, there is a pronunciation to be noted here, I haven’t got it yet, more work to follow…but anyway I’ve got the spelling and meaning and some collocations…and await more on the pron to glue it all together…” This is better than forming the memory firmly around the wrong pronunciation. It’s harder to retro-fit a new pron once the wrong one has been stored.
Work from a (graded) reader
When anyone is reading, and certainly when reading in a second language, it is likely that there is a subvocalized inner voice speaking in the learner’s inner ear, and this voice has a pronunciation. Try this for yourself. This means that even silent reading includes real time rehearsal of pronunciation. A pronunciation of some sort is being rehearsed yet no extra time is required to do it. And the question is, which pron is being rehearsed? Is it the pron of L2, as far as possible,? Or is it L1 pron that is being applied to L2 words? If the teacher does nothing about it then the St may quite naturally default to her L1 pron, There is scope for working on pron in the inner voice and the inner ear. I use this a lot and it has some great advantages too.
In a similar manner writing entails the selecting of words in the head, trying them out in the inner voice, changing them until satisfied, and then writing them down. All of this, including the writing itself is accompanied by an inner voice, which has a pronunciation. Try this for yourself. Once you see this in yourself you can exploit it with your learners.
It’s interesting to reflect that there is no such thing as reading and writing without inner speaking and inner listening, and therefore inner pronunciation. Is that true?