My idli batter worked perfectly all summer and stopped working in October. Same rice, same dal, same ratio, same soak. Flat, dense, sour idlis for about two months before I understood.
It was the temperature of my kitchen, and it is the single largest variable in any fermentation, and almost no recipe mentions it.
What fermentation actually needs
The organisms doing the work — wild yeasts and lactic acid bacteria — have a temperature range in which they are active. Below it they slow dramatically. Above it, different organisms take over and you get sourness without rise.
The comfortable range for idli batter is somewhere around the high twenties in Celsius, which is a normal room temperature in southern India and a warm summer day in most other places.
My kitchen in October is about eighteen degrees. At eighteen degrees the fermentation still happens, and it takes something like twenty hours instead of eight, and I was checking at eight and concluding it had failed.
The signs are not what I thought
I had been judging by volume. Batter should roughly double, so if it had not doubled it had not worked.
Volume is a poor indicator on its own, because a cold ferment produces gas more slowly and some of it escapes before the batter traps it.
Better indicators: the smell, which should be pleasantly sour rather than raw; the texture, which becomes visibly aerated when you stir it; and small bubbles across the surface rather than a few large ones.
A batter that smells right and is full of small bubbles will make good idlis even if it has not visibly doubled.
What worked in a cold kitchen
Several things, in ascending order of effort.
Leaving it longer, which is free and is the answer most of the time. Twenty hours instead of eight.
Finding a warm spot. The top of a fridge is warmer than a counter. An oven with only the light on holds a gentle warmth. Inside a switched-off oven with a bowl of hot water beside it works well.
Using warm water for the grinding, which gives the batter a head start.
Wrapping the bowl in a towel, which does nothing for temperature but does reduce the swing between day and night, and stability seems to matter.
What did not work: adding commercial yeast, which produces rise without the flavour and makes something that is not really idli.
The salt argument
There is a genuine disagreement about whether to salt before or after fermenting, and both camps are confident.
The case for after is that salt inhibits microbial activity, so a cold kitchen needs all the help it can get.
The case for before is that salt suppresses the wrong organisms more than the right ones, producing a cleaner flavour.
In my kitchen, in winter, salting after ferments faster. In summer I salt before because speed is not the constraint and I prefer the taste. That is an empirical position rather than a principled one.
Over-fermentation is real
Once I solved the cold problem I overcorrected and left batter for thirty hours in a warm spot.
The result was aggressively sour, and the idlis had a coarse texture and collapsed slightly after steaming. The structure that traps air had been broken down by too much acid.
Which means it is a window rather than a threshold. Under-fermented is dense and raw-tasting. Over-fermented is sour and collapses. The good zone is wider than it sounds and it is not infinite.
Batter that has gone slightly too far is still fine for dosa, which is more forgiving and where a bit of sourness is welcome. Very little is genuinely wasted.
The general point
This applies to every ferment I have since attempted — dough, yoghurt, pickles, anything.
Recipes are written by people in their own kitchens, at their own temperatures, and the times they give are specific to conditions they did not think to describe.
So the times in any fermentation recipe should be read as approximations from someone else's conditions, and the actual instruction is the description of what it should look and smell like.
Once I started reading recipes that way, a whole category of failures stopped happening.
The grinder makes a difference too
Worth adding because it is the second variable people underestimate.
A wet grinder produces a batter with a texture that a blender does not, largely because a blender heats the batter through friction and a stone grinder does not.
Heat during grinding damages the starches and can partially kill the organisms you are relying on, which is why blender batters sometimes ferment sluggishly for reasons that look mysterious.
If a blender is what you have, the fix is to use cold or iced water and to grind in short bursts with pauses, keeping the batter cool to the touch throughout.
It is not equivalent to a proper grinder and it removes one more reason for an otherwise unexplained failure.