This is to fix BF where tenant migration recipient is failing with OOM while conducting 50 parallel migrations. Apparently only Amazon builds are affected - as a variant make it conditional on Amazon and decrease count from 50 to ~20.

In general I would like to know if we receive a feedback signal from Cloud on near-OOM condition. The proper strategy would be gate-keep concurrent migration on system-wide signals in general, and near-OOM RAM in particular. If any such handler exists please make a comment. If/when such a fix is made, it would be better to scale the migration count back up.