It looks like the decree № 59 was not implemented and replaced by decree № 147 <http://istmat.info/node/35854> which postponed the implementation until July 1, 1919. This is close to the date in tzdata, but it would still make more sense to have transition on a round UTC hour rather than at 21:28:41.

On Tue, Aug 23, 2016 at 11:57 PM, Alexander Belopolsky <alexander.belopolsky@gmail.com> wrote:

On Tue, Aug 23, 2016 at 9:25 PM, Paul Eggert <eggert@cs.ucla.edu> wrote:
>>
>> What's the difference between MSM (Midsummer) and MDST (Moscow Double
>> Summer Time)? Both seem to be +05 = MSK + 2, but got different
>> abbreviations.
>
>
> The former differs from UT by an integer number of hours, the latter does not.

Interesting. It looks like the transition to integer offset from UTC is recorded as happening on Mon Jun 30 21:28:41 1919 UT (from zdump -v output.) However, according to this February 18, 1919 decree № 59 <http://istmat.info/node/35567>, the GMT-based system of 11 zones (belts) went into effect on GMT midnight, April 1, 1919. The decree does not mention Moscow time, but the UTC+2 belt boundaries are defined by the state boundaries on the West and rivers Volga and Don on the east which places Moscow in UTC+2. The decree does not establish DST rules, but allows regional DST shifts by integer number of hours only.