All three turn a Number into a string; they differ in what they count and what they round.
toFixed(digits)gives fixed-point notation with exactlydigitsdigits after the point (0–100, elseRangeError). For values of 10²¹ or more it gives up and returns the normaltoStringform, such as"1e+21".toPrecision(precision)givesprecisionsignificant digits (1–100). It switches to exponent notation when the exponent is below −6 or at leastprecision, so(123456).toPrecision(2)is"1.2e+5".toLocaleString(locales, options)is redefined by ECMA-402: it constructs anIntl.NumberFormatand formats with it. You get grouping separators, the locale's decimal mark, currency, percent and units. Its default for plain numbers is at most three fraction digits.
The 1.005 puzzle: the literal 1.005 has no exact binary representation, so the stored double is 1.00499999999999989…. toFixed and toPrecision work on that exact value: the spec picks the integer n that makes n / 100 − value closest to zero, which is 100, so "1.00". Ties (only possible when the stored value is exactly halfway, like 1.25) go to the larger magnitude.
toLocaleString behaves differently. ECMA-402's ToIntlMathematicalValue first converts the Number with Number::toString, the shortest decimal that round-trips, which is "1.005". It then rounds that decimal with the default roundingMode of "halfExpand", so (1.005).toLocaleString('en-US', { maximumFractionDigits: 2 }) is "1.01".
So the two families can disagree on the same number. For display, prefer Intl.NumberFormat with explicit options; for money, store integer minor units (cents) and format at the edge.